Vous êtes sur la page 1sur 14

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/316744592

The effects of cognitive capacity and gaming


expertise on attention and comprehension:
Working memory and expertise in...

Article in Journal of Computer Assisted Learning January 2017


DOI: 10.1111/jcal.12193

CITATIONS READS

0 31

2 authors:

Yu-Hao Lee Carrie Heeter


University of Florida Michigan State University
26 PUBLICATIONS 239 CITATIONS 95 PUBLICATIONS 1,955 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

RelaxU Meditation App View project

"Movement Meditations" for Virtual Reality View project

All content following this page was uploaded by Yu-Hao Lee on 10 May 2017.

The user has requested enhancement of the downloaded file.


Original article
doi: 10.1111/jcal.12193

bs_bs_banner

The effects of cognitive capacity and gaming


expertise on attention and comprehension
Y.-H. Lee* & C. Heeter
*Department of Telecommunication, University of Florida, Florida, USA
Department of Media and Information, Michigan State University, Michigan, USA

Abstract Educational video games can impose high cognitive demands on its users. Two studies were
conducted to examine the cognitive process involved in playing an educational digital game.
Study 1 examined the effects of users working memory capacity and gaming expertise on
attention and comprehension of the educational messages. The results showed that gaming
experts seem to benet more from having a higher working memory capacity when processing
information from the game. However, gaming experts available working memory did not
predict better comprehension. Instead, non-experts available working memory predicted
better comprehension. Study 2 further examined whether these results were caused by
insufcient working memory allocation or different attention focus between gaming experts
and non-experts. The ndings suggest that gaming experts approach the game differently
from non-experts, focusing on familiar features and overlooking unfamiliar (educational)
information.

Keywords attention, cognitive process, digital game-based learning, expertise, working memory.

Introduction Mere exposure to educational games does not predict


learning outcomes. Only recently have researchers
Digital games are increasingly used for educational
began investigating how learners cognitively process
purposes. Digital games are particularly effective at
information in multimedia environments such as a digital
teaching complex issue through motivating learners to
game (e.g., Lang, 2000; Lang, Zhou, Schwartz, Bolls, &
solve simulated problems (Papastergiou, 2009). Games
Potter, 2000; Mayer & Johnson, 2010; Moreno & Mayer,
encourage players to take on multiple perspectives
2007; Reeves & Thorson, 1986; Vorderer, Knobloch, &
(Peng, Lee, & Heeter, 2010) and to experiment with
Schramm, 2001). Both design features and learners
different solutions in safe environments (De Freitas &
cognitive abilities affect how learners learn from games.
Neumann, 2009; Gee, 2007; Squire, 2011). However,
Certain designs facilitate cognitive processing and
reviews on the effectiveness of digital games for
learning outcomes, while other designs create
learning have been mixed (Kharrazi, Lu, Gharghabi,
unnecessary burden on learners (Mayer & Johnson,
& Coleman, 2012; Kirriemuir & McFarlane, 2004).
2010; Mayer & Moreno, 2003). On the learners side,
Left to their own, learners can sometimes draw
an individuals working memory capacity (WM) limits
incorrect conclusions (Barab, Thomas, Dodge,
the amount of information one can process at any given
Carteaux, & Tuzun, 2005).
time. Learners who possess higher WM or can process
game-related information more efciently may have an
advantage over learners who have lower WM. As an
Accepted: 30 March 2017
Correspondence: Yu-Hao Lee, Department of Telecommunication,
interactive medium, digital games demand a large
University of Florida, Weimer Hall, Gainesville, Florida 32611, USA. amount of cognitive resources from its players. A wide
Email: leeyuhao@jou.u.edu range of cognitive abilities are involved in the process

2017 John Wiley & Sons Ltd Journal of Computer Assisted Learning 1
2 Y.-H. Lee & C. Heeter

of playing a digital game, from visualspatial navigation, certain information while blocking out interference.
perceptual motor skills, accurate timing of responses, as WM is related to many higher-level cognitive abilities
well as short-term and long-term strategic decisions such as comprehension, reasoning and problem-
(Fisch, 2000; Green & Bavelier, 2003; Grodal, 2000; solving (Bull & Scerif, 2001; Daneman & Merikle,
Logie, Baddeley, Donchin, & Sheptak, 1989). Players 1996; Yuan, Steedle, Shavelson, Alonzo, & Oppezzo,
often must process information in various modes 2006), all of which are commonly found in
simultaneously to construct a mental model of the system educational digital games that require players to
(Moreno & Mayer, 2007). simultaneously process information in different
As digital games become more prevalent in sensory modes while constructing their mental model
educational settings, it is important to investigate of the problems.
how learners cognitive abilities affect their learning There are two major paradigms in studying working
from digital games and whether some learners are memory. The rst paradigm focuses on identifying
more likely to benet or become disadvantaged from which conceptual component of working memory is
digital game-based learning. Two experiments were associated with different cognitive tasks, usually
conducted to examine the effects of WM and gaming through selective interference or dual task paradigms
expertise (GE) on attention and comprehension. Study (Baddeley, 2007; Gopher, 2006). The second
1 examined how WM and GE affect attention to the paradigm focus on examining the effects of WM on
game and comprehension of the educational message. individual performance in complex cognitive tasks
Study 2 further used in-game behavioural data to (Engle, 2002; Logie et al., 1989). This approach treats
investigate whether the difference between gaming WM as an individual trait and argues that WM
experts and non-experts was due to different amount primarily inuences performance in complex
of working memory allocated to the task or difference cognitive tasks through the executive control function
in how they allocated their working memory and (e.g., Cowan et al., 2005; Engle, 2002; Kane,
attention. Bleckley, Conway, & Engle, 2001). This study follows
the second paradigm in focusing on the effects of WM
on attention and comprehension.
Theoretical background Attention control is involved in the selection of
information from external stimuli and retrieval of
Working memory capacity in digital games
existing information stored in ones long-term
There is a consensus that humans have a limited WM memory. Individuals with higher WM can retrieve
(Lang, 2000; Mayer & Johnson, 2010; Sweller, 1994; goal-relevant information more efciently than
Van Merrienboer & Sweller, 2005). When processing individuals with lower WM (Barrett, Tugade, & Engle,
information, people continuously draw upon their 2008). The attention control function of working
limited pool(s) of working memory to temporarily memory is also involved in the suppression of
store information and to direct their attention automatic or habitual information processing. Kane
(Baddeley, 2007; Lang, 2000; Moreno & Mayer, and Engle (2003) used a Stroop test to examine
2007). The amount of working memory involved in individual differences in suppressing automatic
a task and how working memory is allocated are response and found that individuals with higher WM
determined by media features and user characteristics made fewer errors on incongruent trials (e.g., the word
(Kalyuga, Ayres, Chandler, & Sweller, 2003; Sweller, RED printed in green ink) than individuals with lower
1994). Working memory is a cognitive system that WM, suggesting that individuals with higher WM are
determines the temporary storage capacity and the more capable of focusing on identifying the colour of
executive control of attention during information the words while suppressing automatic attention to
processing (Baddeley, 2007). The temporary storage reading the words. The goal of this study is to
function stores different modes (visualspatial, investigate how WM and GE affects learners attention
phonological and episodic) of information for and comprehension. Based on theories about working
processing (Baddeley, 2000), while the executive memory, we posit that WM will predict learners
control function selectively allocates attention to attention to the game:

2017 John Wiley & Sons Ltd


Working memory and expertise in games 3

H1. WM will predict more available working memory available working memory when playing an
[attention: measured through shorter secondary task educational game.
reaction time (STRT)] when playing an educational
game. H2. GE will predict more available working memory
(shorter STRT) when playing an educational game.

Expertise and working memory Working memory and comprehension in digital games

Domain-specic expertise is another factor that Educational digital games often embed the educational
interacts with WM to inuence attention and message within its mechanics and narratives. The
comprehension. Studies suggest that domain expertise mechanics of the game refers to the controls, physics,
may facilitate processing of domain-related resource and limitations that afford the interactive
information, but not unrelated or general information. experience. Narrative refers to information about
For example, individuals who were knowledgeable characters, events, environment and relations. As an
about baseball could recall a story about a baseball interactive medium, digital games require players active
game better than individuals who were low in baseball input and decision making to develop a narrative
expertise (Spilich, Vesonder, Chiesi, & Voss, 1979). (Klimmt & Hartmann, 2006). Therefore, the rst task
Another study found that baseball expertise facilitated for a game player is to learn the game mechanics. Only
memory of baseball-related words, but not neutral ones after the player masters the mechanics can the narrative
(Fincher-Kiefer, Post, Greene, & Voss, 1988). Domain be constructed through the players interaction with the
experts have domain-specic schema stored in their game (Lee & Faber, 2007). The educational message
memory to assist them in processing domain-related refers to the underlying concepts and knowledge that
information. When an expert chess player encounters the source (game designer or instructor) intends to
a chess set, they do not process each chess piece as communicate through an educational game. During
separate pieces of information. Instead, they can search gameplay, the players working memory is allocated to
their long-term memory and retrieve a similar pattern to learning and processing the mechanics rst; then the
facilitate their information processing. This expertise residual available working memory is allocated to
advantage disappeared when a board of randomly processing the narrative and educational message (Lee
positioned chess pieces was presented (Chase & Simon, & Faber, 2007). If the mechanics are too demanding,
1973). Domain experts may have an advantage when there will be less working memory available for
the information can be connected to their existing processing the narrative and educational message (Elson,
schema. Breuer, Ivory, & Quandt, 2014; Fisch, 2000). Working
There are two types of expertise when playing memory demand is lower if the educational message is
educational games expertise in the educational closely embedded in the mechanics or narrative but
content (content expertise) and expertise in playing increases if the educational message is distant to the
digital games (GE). Content expertise helps players mechanic and narrative (Fisch, 2000). Not all digital
understand the narrative of the message, while GE games have explicit narratives, but all digital games have
helps players identify goal-relevant and irrelevant mechanics, and the mechanics take precedence over
information on the basis of their experience with narrative and educational messages when competing
similar games. Studies have found that gaming for the players working memory.
experts could track objects moving at greater speeds, Players must process many information regarding
were more accurate in a visual short-term memory mechanics, narratives and educational message in an
test, performed task switch more quickly, as well as educational game. Because gaming experts are
monitored and evaluated rotating objects faster and familiar with similar game mechanics, they may have
more accurately (Boot, Kramer, Simons, Fabiani, & more available working memory to help them
Gratton, 2004). Because gaming experts can process identify and process goal-related information in a
information in games more efciently, we game. As a result, they may benet more from
hypothesize that players with higher GE have more having higher WM to sustain their attention on the

2017 John Wiley & Sons Ltd


4 Y.-H. Lee & C. Heeter

goal-related information and suppress automatic Study 1


responses. However, gaming experts with lower
Method
WM may not have the capacity to control their
automatic responses and instead focus on a
Participants
narrower subset of the familiar information and feel
cognitively strained. In other words, will gaming Undergraduate participants (n = 93) were recruited from
experts benet from having more WM? We a large Midwestern university. The students received
extra credits in their courses as incentives for
hypothesize that WM will moderate the effect of
GE on attention (H3). participating in the experiment. The mean age was
21.05 years (SD = 2.40) and ranged from 19 to 27 years.
There were slightly more men (54.8%) than women
H3. There will be an interaction effect between WM and
GE on available working memory (attention). (45.2%) in the study.
Participants with higher WM and GE will have more
available working memory than participants with high Stimuli and educational message
GE but lower WM. The educational game used in this study was The
ReDistricting Game, a game about political redistricting
Although gaming experts may have an advantage in and the problem of gerrymandering. Gerrymandering is
allocating their working memory to processing the act of manipulating voting districts to gain unfair
information in an educational game, gaming experts advantage for certain political parties or interest groups
may not always have an advantage in learning the in an election. In the game, players play as a map maker
educational message. In some contexts, their existing who draws voting districts. Mission 1 is a basic tutorial
schema may mislead them to focus on familiar about the controls and mechanics. Mission 2 is where
information while overlooking new, important the educational message is introduced about partisan
information. Ricks, Turley-Ames, and Wiley (2007) gerrymandering, which requires players to draw voting
tested whether experts and non-experts were able to districts to favour a political party. The games goal
detect questions designed to mislead experts into was to teach players about how redistricting works and
automatic processing and construct false answers. how it could easily be manipulated to undermine
Their ndings showed that when the task appeared democracy. The ReDistricting Game was launched in
to be within the experts expertise domain, experts 2006 and was used by congressmen, senators and
quickly activated habitual processing and failed to lobbyists to bring awareness to gerrymandering and push
detect that their answer was incorrect. In comparison, for legislative reforms. The game has also been adopted
because non-experts do not have existing schema to in civic education courses in schools and has shown to
guide them, they were more likely to process the be effective in raising students interest and
question cautiously, process more information and understanding of the redistricting process and the
answer correctly. When the task is structurally similar problems of gerrymandering (Juckett & Feinberg,
to the experts expertise domain, experts may search 2010; Lerner, 2014).
for familiar information while overlooking the Like many educational games, The ReDistricting
unfamiliar educational message as irrelevant Game was designed to highlight a problem. The explicit
information. As for non-experts, because they do not mechanics of the game involves manipulating election
have existing schemas, they are more likely to process results through gerrymandering, but the underlying
the new information altogether, which can potentially educational message is that gerrymandering is a
lead to better comprehension if they have sufcient problematic practice and can undermine democracy.
WM. Because we do not know if gaming experts will The educational message is delivered through feedback
have an advantage or be misguided by their existing from non-player characters and additional hyperlinks
schema, we pose a research question: will gaming that provide additional information about
experts attention to the game predict better gerrymandering. This type of design in which the
comprehension when compared with non-experts? mechanics are not closely aligned with the educational
(RQ1). message requires more working memory from the player

2017 John Wiley & Sons Ltd


Working memory and expertise in games 5

to process (Fisch, 2000). Players must rst learn the memory while processing the math calculations. Three
mechanics of the game, practise gerrymandering and sets of four questions each (total of 12 questions) were
then learn from the feedback and links that the intended used to measure WM. WM score was calculated by the
educational message was that gerrymandering is number of correct answers ranging from 0 to 12.
problematic. A case study by Juckett and Feinberg Participants were given 7 s to solve each question before
(2010) found that while most students enjoyed learning the page automatically moved on; this prevented
about gerrymandering through this game, students with participants from intentionally rehearsing the words. An
less gaming experience were frustrated and some 80% accuracy criterion on the math test was used to
preferred the traditional paper-and-pencil approach of ensure that participants were processing the math
learning the game. Such nding suggests that learners question while storing the words in their working
with varying degrees of GE may have different responses memory instead of only focusing on the words. The
to digital game-based learning. While not all educational mean AOSPAN score was 5.98 (SD = 3.66).
games are designed with distant mechanics and
educational messages, using a game with high working
memory demands allows this study to tease apart the Dependent variables
different effects of WM and GE on players available Available working memory (measured by secondary task
working memory and comprehension. reaction time). Available working memory was measured
through STRT. STRT is a measurement commonly used
Operational measures to measure attention through measuring available
working memory (Lang, Bradley, Park, Shin, & Chung,
Gaming expertise. Gaming expertise was measured by a 2006). STRT was measured by instructing participants to
series of 7-point questions in the pretest survey. focus on a primary task (playing the game). During the
Participants were asked about self-reported familiarity primary task, the participants were given a secondary
and expertise with a list of different genres including task of responding (pressing a button) to a signal called
strategy, puzzle, action, sports, role-playing and the secondary task probe. The latency between the probe
adventure. Because the game used in this study is a and the reaction is the STRT. A shorter STRT suggests
strategy simulation game, the mean score of familiarity more available working memory as it indicates the
and expertise with strategy and simulation games (four participant has sufcient WM to process the two tasks
items) was used as the measurement of GE. The mean simultaneously. A longer STRT suggests less available
score was 3.48 (SD = 1.61). The items were reliable at working memory because the participant cannot respond
= 0.70. Because previous studies suggest that there fast enough to the secondary task probe when processing
may be differences in genre preference between genders the primary task.
(Lucas & Sherry, 2004), we conducted an independent In this study, the secondary task probe was a yellow
sample t-test to see if there were gender differences on box that appeared at random times and places around
GE. The result suggests that there was no signicant the game screen; participants were instructed to press
gender difference, t(90) = 1.13, p = 0.262. the space bar key as fast as you can when they the box
Working memory capacity. Working memory capacity was
appeared. On average, participants were exposed to
measured using the automatic operation span test
6.31 secondary probes. Because not all the participants
(AOSPAN; Unsworth, Heitz, Schrock, & Engle, 2005).
received an equal number of secondary probes, to allow
Participants were instructed to solve math operations
comparison between participants, only the rst four
while trying to remember a set of words. For example,
STRTs of mission 2 (which included the educational
a question looks like this:
message) were included in the analyses. Missed
Is 1 28 15? Yes=No Truck responses (i.e., no response after 5000 ms from the onset)
were coded as missing data. The mean STRT was
The string of math operation and words appears one at 1090.45 ms (or 1.09 s, SD = 489.65 ms), ranging from
a time. After each set of questions, participants were 461 to 3672 ms.
asked to identify the words in the correct order. This task Comprehension. Comprehension was measured with 11
required participants to hold the words in their working multiple-choice questions that tested the participants

2017 John Wiley & Sons Ltd


6 Y.-H. Lee & C. Heeter

comprehension of the redistricting process. Four items from intentionally neglecting the secondary task, they
measured general understanding of gerrymandering, four were told that if they performed poorly on the secondary
items pertained to redistricting principles and process, task, they need to redo the task.
and three measured knowledge about solutions presented The participants played mission 1 of the game
in the game. Each correct answer was given 1 point. followed by mission 2 until they complete mission 2 or
Thus, the comprehension score ranges from 0 to 11. for 30 min, whichever happens rst. Thirty minutes of
The mean comprehension score was 4.84 (out of 11, gameplay was determined to be appropriate because the
SD = 2.31) ranging from 0 to 11. The reliability for the game was designed to be completed within one class
comprehension items had a Cronbachs of 0.67, which period. Mission 1 was a tutorial of the game mechanics
is acceptable because the questions were not and narratives. Mission 2 introduced the core educational
unidimensional but measured three different types of message about the problems of gerrymandering. The
knowledge about the redistricting process. secondary task probe appeared at random intervals with
at least 30 s between signals. Each signal remained on
the screen for 5 s. If the participant did not respond
Control variables
within the 5 s, the system recorded the response as
Existing content knowledge about the redistricting
missing data. After the game, participants lled out a
process may inuence the educational comprehension
post-test questionnaire measuring their enjoyment of
scores. Two 7-point questions were used to measure the
the game, their perceived challenge level and their self-
participants self-assessment of content knowledge: I
reported effort. These questions acted as distractors
am very familiar with the redistricting process and I
before the comprehension test. This helps ensure that
consider myself an expert in how gerrymandering
the comprehension test measured information stored in
works. The questions were designed so that they do
long-term memory. Comprehension was measured by
not prime the participants to pay attention to specic
11 multiple-choice questions about redistricting
parts of the game. The two items were reliable at
and gerrymandering. Lastly, participants answered
r = 0.74, the mean for content knowledge was 2.86
demographic questions including age, sex and race.
(SD = 1.65). The score was skewed towards the lower
end, which suggests that participants had low existing
content knowledge. Results
Sample
Experiment procedure
The study was conducted in a computer lab at a large Two participants were removed because they did not
Midwestern university. After agreeing to the informed meet the 80% math accuracy criterion on the AOSPAN
consent, the participant rst answered a short test. Three more were removed because they did not
questionnaire measuring their political involvement, respond to any secondary task probes when playing the
their political efcacy and content knowledge. After the game. As a result, 88 participants were included in the
pretest questionnaires, participants read a detailed analyses.
instruction about the AOSPAN test. They were given
three practice questions to familiarize themselves with
Hypotheses testing
the AOSPAN test. After the practice questions, they took
the actual AOSPAN test consisting of 12 questions. H1 posited that WM would positively predict shorter
Following the AOSPAN test, the participants read an STRT when playing the game; H2 posited that GE would
instruction reminding them to pay attention to the game positively predict shorter STRT when playing the game.
because they would later be tested about the contents. H3 posited that there would be an interaction effect
The instructions were given to increase motivation and between WM and GE on STRT. We tested the three
to frame the game as a learning experience. After reading hypotheses through a hierarchical regression analysis.
the game instructions, the participants read an instruction The mean-centred WM (AOSPAN) score and mean-
about the secondary task probe and were instructed to centred GE score were entered into the rst block as
respond as fast as possible. To prevent participants independent variables; the interaction term between

2017 John Wiley & Sons Ltd


Working memory and expertise in games 7

WM and GE was entered into the second block to test signicant, p = 0.006. The simple slope analysis results
interaction effects. The STRT score was entered as the indicate that when participants had higher WM and
dependent variable. higher GE, their STRT was signicantly shorter,
The analysis result showed that the overall model was suggesting more available working memory. However,
signicant, F(2, 85) = 2.30, p = 0.041, adjusted participants with high GE but lower WM performed
R2 = 0.08. There was no signicant main effect for signicantly slower in STRT, suggesting that they did
WM ( = 0.11, t = 1.02, p = 0.150) and no signicant not have sufcient working memory to process the game
main effect for GE ( = 0.15, t = 1.21, p = 0.165). The and the secondary task at the same time. The results were
result was not consistent with H1 or H2, which suggests consistent with H3, gaming experts benetted from
that WM or GE alone did not predict available working higher WM but suffered when they have lower WM. See
memory (STRT) when playing an educational game. Figure 1 for simple slope analysis.
However, there was a signicant interaction effect RQ1 asked: Will gaming experts attention to the
between WM and GE on available working memory game predict better comprehension when compared with
( = 0.24, t = 2.22, p = 0.014). In order to further non-experts? A hierarchical regression was conducted to
examine the interaction effect, we conducted a simple examine RQ1, content knowledge and WM were entered
slope analysis using the PROCESS plug-in for SPSS by in the rst block as control variables; the mean-centred
Hayes (2012). The standardized for participants 1 SD GE and STRT were entered in the second block, and
below the average GE was 0.12 and was not the GE STRT interaction term was entered in the third
signicant, p = 0.233. The standardized for participants block. Comprehension score was entered as the
with average GE was 0.24 but was signicant, dependent variable. Analysis results showed that the
p = 0.014. The standardized for participants 1 SD overall model was signicant, F(5, 82) = 3.26,
above the average GE was 0.39 and was also p = 0.010, adjusted R2 = 0.12. There was a signicant

Figure 1 Simple Slope Comparison for GE WM on Available Working Memory. GE, gaming expertise; WM, working memory capacity

2017 John Wiley & Sons Ltd


8 Y.-H. Lee & C. Heeter

main effect for GE ( = 2.42, p = 0.009), a signicant comprehension of the educational message. There are at
main effect for available working memory ( = 0.31, least two potential explanations: (1) insufcient working
t = 2.10, p = 0.019) and a signicant interaction effect memory allocation or (2) misallocation of working
( = 1.97, p = 0.026). A simple slope analysis was memory.
conducted to interpret the interaction effect. The The rst possibility may be due to insufcient
standardized for participants 1 SD below the average working memory allocation. That is, gaming experts
GE was 0.286 and was signicant, p = 0.044. The thought the task was easy to them and allocated very
standardized for participants with mean GE was little working memory to processing the task, while
0.680 and was also signicant, p = 0.014. The non-experts allocated more working memory to the
standardized for participants 1 SD above the average task because they expect to be challenged. Research
GE was 0.120 but was not signicant, p = 0.150. The in textual comprehension has shown that experts can
ndings suggest that when controlling for the effects of perform worse when reading simple text but benetted
WM and content knowledge, participants with higher from reading an unstructured text because they are
GE scored higher in comprehension in general. GE also challenged to use existing schema to ll in the gaps
moderated the effect of available working memory on (McNamara, Kintsch, Songer, & Kintsch, 1996). The
comprehension. Surprisingly, the simple slope analysis second possible explanation is misallocation of
indicated that available working memory (attention) working memory. Gaming experts might have been
was not a signicant predictor of comprehension for focused on retrieval of their existing schema or on
gaming experts, but it was a signicant predictor of identifying familiar information but not on processing
comprehension for non-experts. the new, unfamiliar information. In other words,
because they recognize the game mechanics, they used
their existing schema to search for similar mental
Study 1 discussion
models but overlooked the unfamiliar educational
Study 1 found that the learners WM moderated the message. Study 2 was conducted to test these two
effect of GE on their available working memory potential explanations.
during gameplay. Specically, there was a signicant
difference between gaming experts. Gaming experts
Study 2
with higher WM had more available working memory
than gaming experts with lower WM, possibly because To further examine whether the results from study 1 were
of having the expertise and cognitive capacity to caused by insufcient working memory allocation or
identify relevant information, navigate the game and misplaced attention, study 2 replicated study 1 but added
process information communicated through a game a more difcult condition to test the insufcient working
more efciently. memory allocation explanation. Game-behavioural data
While GE predicted better comprehension, we found were tracked to examine the misallocated attention
that the available working memory among gaming explanation. A more difcult condition was included
experts did not predict comprehension. Instead, available because studies in text comprehension have found that
working memory among non-experts predicted better experts may benet more from reading more difcult
comprehension. How do we interpret this result? The incohesive text rather than a standard, familiar
ndings were consistent with literature that showed structure. This is because the incohesive text contains
experts benetted more from having higher WM because gaps that force experts to ll with their existing schema,
they can identify relevant information and process the which improves their attention and comprehension
information more efciently (Chi, Glaser, & Farr, (McNamara et al., 1996; Oreilly & McNamara, 2007).
1988). However, at the same time, experts can also If the difference between gaming experts and non-
over-rely on their existing schema when they recognize experts was due to insufcient working memory
the task as within their expertise domains, causing them allocation, we should see the gaming experts performing
to focus on familiar information and overlook new better in comprehension (H4) in the more incohesive
information (Ricks et al., 2007). It is unclear why gaming condition when compared with the standard condition
experts available working memory did not predict better used in study 1.

2017 John Wiley & Sons Ltd


Working memory and expertise in games 9

Another potential explanation for the different In-game behaviour measurements


performance between gaming experts and non-experts Time in instruction screen. The amount of time that the
is that they were focusing on different aspects of the players spent on the instruction screen is measured in
game. We hypothesize that gaming experts would
seconds.
approach the game differently than non-experts. To
examine this explanation, in study 2, we tracked
in-game behaviour data from the server hosting the Number of help functions used
game to examine whether the gaming experts and The game contained a help icon that was constantly
non-experts approached the game differently. We available during gameplay for players to reference if they
hypothesize that because gaming experts are more need help with the controls and goals. The use of this
likely to rely on their existing schema to guide them, function is not mandatory. Thus, it is a measure of how
they would spend less time reading instructions (H5), many times the players actively sought additional
seek less help (H6), and therefore make more failed support.
attempts (H7). However, because they are focused on Number of failed attempts. When the players have
their schema, they will spend less time reading completed drawing voting districts, the player can submit
corrective feedback (H8). the map for review. The district map must meet certain
criteria mandated by redistricting laws. Cautious players
who are aware of the criteria will only make a submission
Method when they are condent that the districts meet the
requirements. In comparison, players who are not paying
attention to regulations will make hastier submissions
Participants
and more failed attempts.
A total of 97 undergraduate participants were recruited
Number of feedback read. Formative feedback was given to
for study 2 from several courses at a large Midwestern
players when they make changes to the map; the
university. The mean age was 21.78 years (SD = 1.63)
feedback tells the players what they are doing right and
with equal numbers of male and female participants
how they can improve. The feedback messages do not
(n = 48 each, one participant did not answer).
appear unless the players actively click on the feedback
button. Therefore, this measure tracks how many times
Procedures the players actively sought feedback to learn about their
The experiment procedures were identical to study 1 with performance and correct their decisions.
the addition of a more challenging, incohesive condition
in which the participants were instructed to play mission
Study 2 results
2 of The ReDistricting Game without playing through
the tutorials in mission 1. To reduce the possibility that H4 tested whether gaming experts allocated insufcient
outcome difference was due to exposure time, the working memory to the game. If this explanation was
participants in the incohesive condition were instructed supported, we should see higher comprehension in
to play mission 2 for 5 min rst to familiarize with the incohesive condition compared with the standard
interface, then replay mission 2 again until they complete condition. Following the procedures of study 1, we
the mission or for 30 min. The standard condition was conducted a hierarchical regression analysis with prior
identical to study 1. Also, we tracked in-game content knowledge in the rst block as control variables,
behavioural data from the server for all the participants. mean-centred WM, GE and the dummy-coded game
To ensure that the manipulation of the incohesive conditions (0 = standard, 1 = incohesive) in the second
condition made the game more challenging, we asked block; two-way interaction terms were entered into the
the participants to rate the games challenge level using third block; and the three-way interaction term was
a 7-point scale. Manipulation check showed that the entered into the fourth block. Comprehension was the
incohesive condition was rated signicantly more dependent variable (DV). The results showed that there
difcult than the standard condition, t(96) = 2.72, was no signicant main effect for the conditions
p = 0.008. ( = 0.25, p = 0.148) nor for the interaction between

2017 John Wiley & Sons Ltd


10 Y.-H. Lee & C. Heeter

GE and conditions ( = 0.23, p = 0.185). The gaming interaction effect for WM GE ( = 0.09, p = 0.197).
experts did not perform better in comprehension when The data were partially consistent with H8; gaming
assigned to the incohesive condition. Therefore, the data experts read fewer feedback, perhaps because they had
did not support the insufcient working memory a false sense of understanding and did not need to read
allocation explanation. additional information. WM did not moderate the
H5 through H8 examined the second explanation that relationship between GE and number of feedback read.
gaming experts were focusing on familiar mechanics, but
overlooking the educational message. To test the
Study 2 discussion
hypotheses, four separate regression analyses were
conducted, total game time was entered into the rst The goal of study 2 was to further examine whether the
block as a control variable, centred WM and GE were ndings from study 1 were caused by insufcient
entered into the second block as independent variables working memory allocated or misallocated attention
and the interaction term was entered into the third block. during the game. The ndings suggest that the difference
Time spent in instructions (H5), number of help function between gaming experts and non-experts was not caused
used (H6), number of failed attempts (H7) and number of by insufcient working memory allocation but caused by
feedback read (H8) were the dependent variables. misallocation of attention possibly due to over-reliance
For the analysis regarding time spent in instructions on existing schema. Gaming experts are more likely to
(H5), there were no signicant main effects for WM approach the game with a sense of understanding that
( = 0.11, p = 0.177) or GE ( = 0.14, p = 0.093), there leads them to search for familiar information that are
was also no signicant interaction effect for WM GE coherent with their schema, overlooking important new
( = 0.07, p = 0.268). The data were not consistent with information such as the educational messages.
H5. There was no signicant difference between players
with different WM and GE on time spent in instructions.
General discussion
For the analysis regarding number of help function
(H6), there was no signicant main effect for WM As a complex problem-solving task, digital games can
( = 0.03, p = 0.399) or GE ( = 0.08, p = 0.229). impose high cognitive demands on players, yet few
However, there was a signicant interaction effect for studies have investigated how people cognitively process
WM GE ( = 0.21, p = 0.022). Gaming experts with information through digital games. The current study
higher WM used fewer help functions, perhaps because provides important insights that can advance our
they have a false sense of understanding and do not think theoretical understanding of the cognitive processes
they need help, and their higher WM facilitated their involved in digital game-based learning. Based on the
retrieval of existing schema. Non-experts with higher ndings, we can draw several important insights: (1)
WM used the most help functions, perhaps because they both WM and GE inuenced available working memory
do not know what to focus on, yet they have the capacity and comprehension, but the relationship between the
to process more new information. constructs are more complex than the common belief that
For the number of failed attempts (H7), there was no having higher WM and GE is always benecial to
signicant main effect for WM ( = 0.08, p = 0.253) learning from a digital game. (2) Gaming experts are
or GE ( = 0.08, p = 0.249). However, there was a more likely to follow existing mental models than non-
signicant interaction effect for WM GE ( = 0.17, experts. Researchers and practitioners need to not only
p = 0.048). Gaming experts with lower WM made more consider the amount of working memory demand that a
failed attempts, perhaps because they were more likely to game imposes on the learners but also examine where
process the information according to their existing learners allocate their attention and design the game to
schema and make rash decisions without considering support players so that they will notice and process the
all available information. underlying educational messages.
For the number of feedback read (H8), there was no The current study contributes to the body of studies
signicant main effect for WM ( = 0.01, p = 0.451), that showed expertise as a strong predictor of complex
but there was a signicant main effect for GE cognitive tasks (Hambrick & Engle, 2002; Yuan et al.,
( = 0.22, p = 0.018). However, there is no signicant 2006). The ndings suggest that individuals with higher

2017 John Wiley & Sons Ltd


Working memory and expertise in games 11

GE benetted more from higher WM in processing behavioural data or eye tracking to provide a richer
information from a game. However, the gaming experts picture of not only how much working memory is
available working memory did not predict compre- allocated but also where it is being allocated.
hension. Instead, available working memory from non- This study has several limitations. First, the
experts predicted better comprehension. Consistent participants only played the game once without
with ndings from Hambrick and Engle (2002), the additional guidance from an instructor. This may have
results suggest that WM and GE inuence limited the effectiveness of the game in improving
comprehension as separate predictors. In an comprehension. Second, the results from this study are
interactive media, GE not only facilitates the amount based on a single game. The ndings from this study
of working memory allocated to the task, but more can only be generalized to similar game structures in
importantly, expertise inuenced how individuals which the educational message is distant from the
approach the task. The results from study 2 mechanics and demand high WM to process. Future
suggested that gaming experts likely relied on their studies should replicate this study in games that embed
existing schema when navigating the game, causing the educational message closely with the mechanics to
them to overlook important yet unfamiliar educational investigate if gaming experts will have an advantage over
information, whereas non-experts approached the non-experts because of their focused attention. Third, the
game with a fresh set of eyes and processed the in-game behaviour measures used in study 2 can only
information more cautiously. If the educational infer what the participants were doing; they cannot tell
message is closely integrated into familiar game us why the participants behaved this way, while the
mechanics, perhaps the gaming experts may be able results suggest that gaming experts with higher WM used
to process the information more efciently, but when less help functions, made more mistakes and clicked on
the educational message is not embedded in familiar fewer feedback. This could be caused by their false sense
mechanics, gaming experts can be misled by their of understanding because of having existing schema, or it
schema and overlook the important messages. From could be caused by lack of interest in the game or the
a design perspective, gaming experts may benet if desire to get through the study as soon as possible. More
the educational message is integrated into familiar studies are needed to parse out these possible
game mechanics, but they may suffer from misplaced explanations. Finally, there may be other individual
attention if the messages are not integrated into learner factors that inuence their response to digital
familiar mechanics. game-based learning, such as their reading ability. Future
Most previous studies on cognitive processing of studies should take these individual differences into
mediated message have focused on non-interactive, account.
linear medium such as the television and radio (e.g. Lang For game developers who often seek to maximize
et al., 2000; Lang et al., 2006), a few studies examined player immersion, the nding that gaming experts may
structural features on interactive websites (Lang, Borse, block out unfamiliar information and focus narrowly on
Wise, & David, 2002; Moreno & Mayer, 2007; Wise, familiar information is similar to the concept of ow
Bolls, & Schaefer, 2008). Linear media allowed the (Csikszentmihalyi, 1991; Nakamura & Csikszentmihalyi,
researchers to focus on characteristics of the mediated 2002), in which one is fully immersed in a task that they
message because all users were exposed to the same block out all external distractions. If this is the case,
amount of information and approached the message in when the educational message is not embedded within
the same order. But in complex interactive media such their familiar mechanics, it is possible that ow might
as digital games, player characteristics will determine be detrimental to the comprehension of the intended
how players approach and process the information. This message as more digital games are used as educational
adds another dimension to understanding the cognitive tools for providing learners the opportunity to explore
processing of mediated messages in interactive complex problems through experiential learning. We
multimedia. When studying an interactive media such need to understand how learners approach and process
as digital games, it is important to include measures of information in digital games so that educators and
where the cognitive resources are allocated. Future developers can design better curriculum and game
studies could incorporate measurements such as systems that support different learner needs.

2017 John Wiley & Sons Ltd


12 Y.-H. Lee & C. Heeter

References Fincher-Kiefer, R., Post, T. A., Greene, T. R., & Voss, J. F.


(1988). On the role of prior knowledge and task demands
Baddeley, A. (2000). The episodic buffer: A new component of
in the processing of text. Journal of Memory and Language,
working memory? Trends in Cognitive Sciences, 4, 417423.
27, 416428. https://doi.org/10.1016/0749-596X(88)90065-4.
https://doi.org/10.1016/S1364-6613(00)01538-2.
Fisch, S. M. (2000). A capacity model of childrens
Baddeley, A. (2007). Working memory, thought, and action. comprehension of educational content on television.
Oxford University Press.
Media Psychology, 2, 6391. https://doi.org/10.1207/
Barab, S., Thomas, M., Dodge, T., Carteaux, R., & Tuzun, H. S1532785XMEP0201_4.
(2005). Making learning fun: Quest Atlantis, a game without Gee, J. P. (2007). What video games have to teach us about
guns. Educational Technology Research and Development, learning and literacy: Revised and updated edition. New
53, 86107. https://doi.org/10.1007/BF02504859. York: Macmillan.
Barrett, L. F., Tugade, M. M., & Engle, R. W. (2004). Gopher, D. (2006). Control processes in the formation of
Individual differences in working memory capacity and task units. Psychological science around the world, 2,
dual-process theories of the mind. Psychological Bulletin, 385404.
130, 553573. Green, C. S., & Bavelier, D. (2003). Action video game
Boot, W. R., Kramer, A. F., Simons, D. J., Fabiani, M., & modies visual selective attention. Nature, 423, 534537.
Gratton, G. (2008). The effects of video game playing on https://doi.org/10.1038/nature01647.
attention, memory, and executive control. Acta Grodal, T. (2000). Video games and the pleasures of control. In
Psychologica, 129, 387398. https://doi.org/10.1016/j. D. Zillman & P. Vorderer (Eds.). Media entertainment: The
actpsy.2008.09.005. psychology of its appeal (pp. 197213). Mahwah, NJ:
Bull, R., & Scerif, G. (2001). Executive functioning as a predictor Lawrence Erlbaum.
of childrens mathematics ability: Inhibition, switching, and Hambrick, D. Z., & Engle, R. W. (2002). Effects of domain
working memory. Developmental Neuropsychology, 19, knowledge, working memory capacity, and age on cognitive
273293. https://doi.org/10.1207/S15326942DN1903_3. performance: An investigation of the knowledge-is-power
Chase, W. G., & Simon, H. A. (1973). Perception in chess. hypothesis. Cognitive Psychology, 44, 339387. https://doi.
Cognitive Psychology, 4, 5581. https://doi.org/10.1016/ org/10.1006/cogp.2001.0769.
0010-0285(73)90004-2. Hayes, A. F. (2012). PROCESS: A versatile computational tool
Chi, M. T., Glaser, R., & Farr, M. J. (Eds.). (1988). The nature for observed variable mediation, moderation, and conditional
of expertise. Hillsdale, NJ: Erlbaum. process modeling.
Cowan, N., Elliott, E. M., Scott Saults, J., Morey, C. C., Juckett, E., & Feinberg, J. (2010). The redistricting game:
Mattox, S., Hismjatullina, A., & Conway, A. R. (2005). On Teaching congressional gerrymandering through an online
the capacity of attention: Its estimation and its role in simulation game. Social Education, 74, 274278.
working memory and cognitive aptitudes. Cognitive Kalyuga, S., Ayres, P., Chandler, P., & Sweller, J. (2003). The
Psychology, 51, 42100. https://doi.org/10.1016/j. expertise reversal effect. Educational Psychologist, 38,
cogpsych.2004.12.001. 2331. https://doi.org/10.1207/S15326985EP3801_4.
Csikszentmihalyi, M. (1991). Flow: The psychology of optimal Kane, M. J., Bleckley, M. K., Conway, A. R., & Engle, R. W.
experience. New York: Harper Collins. (2001). A controlled-attention view of working-memory
Daneman, M., & Merikle, P. M. (1996). Working memory and capacity. Journal of Experimental Psychology: General,
language comprehension: A meta-analysis. Psychonomic Bulletin 130, 169. https://doi.org/10.1037/0096-3445.130.2.169.
& Review, 3, 422433. https://doi.org/10.3758/BF03214546. Kane, M. J., & Engle, R. W. (2003). Working-memory capacity
De Freitas, S., & Neumann, T. (2009). The use of exploratory and the control of attention: The contributions of goal
learning for supporting immersive learning in virtual neglect, response competition, and task set to Stroop
environments. Computers & Education, 52, 343352. interference. Journal of Experimental Psychology: General,
https://doi.org/10.1016/j.compedu.2008.09.010. 132, 47. https://doi.org/10.1037/0096-3445.132.1.47.
Elson, M., Breuer, J., Ivory, J. D., & Quandt, T. (2014). More Kharrazi, H., Lu, A. S., Gharghabi, F., & Coleman, W. (2012).
than stories with buttons: Narrative, mechanics, and context A scoping review of health game research: Past, present, and
as determinants of player experience in digital games. future. Games for health: Research, Development, and
Journal of Communication, 64, 521542. https://doi.org/ Clinical Applications, 1, 153164. https://doi.org/10.1089/
10.1111/jcom.12096. g4h.2012.0011.
Engle, R. W. (2002). Working memory capacity as executive Kirriemuir, J., & McFarlane, A. (2004). Literature review in games
attention. Current Directions in Psychological Science, 11, and learning. (pp. 140) Retrieved from http://www.futurelab.
1923. https://doi.org/10.1111/1467-8721.00160. org.uk/resources/documents/lit_reviews/Games_Review.pdf.

2017 John Wiley & Sons Ltd


Working memory and expertise in games 13

Klimmt, C., & Hartmann, T. (2006). Effectance, self-efcacy, Oreilly, T., & McNamara, D. S. (2007). Reversing the
and the motivation to play video games. In P. Vorderer & reverse cohesion effect: Good texts can be better for strategic,
J. Bryant (Eds.), Playing video games: Motives, responses, high-knowledge readers. Discourse processes, 43, 121152.
and consequences (pp. 133145). Mahwah, NJ, US: https://doi.org/10.1080/01638530709336895.
Lawrence Erlbaum Associates Publishers. Papastergiou, M. (2009). Exploring the potential of computer
Lang, A. (2000). The limited capacity model of mediated and video games for health and physical education: A
message processing. Journal of Communication, 50, literature review. Computers & Education, 53, 603622.
4670. https://doi.org/10.1111/j.1460-2466.2000.tb02833.x. https://doi.org/10.1016/j.compedu.2009.04.001.
Lang, A., Borse, J., Wise, K., & David, P. (2002). Captured by the Peng, W., Lee, M., & Heeter, C. (2010). The effects of a serious
World Wide Web orienting to structural and content features of game on role-taking and willingness to help. Journal of
computer-presented information. Communication Research, Communication, 60, 723742. https://doi.org/10.1111/
29, 215245. https://doi.org/10.1177/0093650202029003001. j.1460-2466.2010.01511.x.
Lang, A., Bradley, S. D., Park, B., Shin, M., & Chung, Y. Reeves, B., & Thorson, E. (1986). Watching television:
(2006). Parsing the resource pie: Using STRTs to measure Experiments on the viewing process. Communication Research,
attention to mediated messages. Media Psychology, 8, 13, 343361. https://doi.org/10.1177/009365086013003004.
369394. https://doi.org/10.1207/s1532785xmep0804_3. Ricks, T. R., Turley-Ames, K. J., & Wiley, J. (2007). Effects of
Lang, A., Zhou, S., Schwartz, N., Bolls, P. D., & Potter, R. F. working memory capacity on mental set due to domain
(2000). The effects of edits on arousal, attention, and knowledge. Memory & Cognition, 35, 14561462. https://
memory for television messages: When an edit is an edit doi.org/10.3758/BF03193615.
can an edit be too much? Journal of Broadcasting & Spilich, G. J., Vesonder, G. T., Chiesi, H. L., & Voss, J. F.
Electronic Media, 44, 94109. https://doi.org/10.1207/ (1979). Text processing of domain-related information for
s15506878jobem4401_7. individuals with high and low domain knowledge. Journal
Lee, M., & Faber, R. J. (2007). Effects of product placement in of Verbal Learning and Verbal Behavior, 18, 275290.
on-line games on brand memory: A perspective of the https://doi.org/10.1016/S0022-5371(79)90155-5.
limited-capacity model of attention. Journal of Advertising, Squire, K. (2011). Video games and learning: teaching and
36, 7590. https://doi.org/10.2753/JOA0091-3367360406. participatory culture in the digital age. Technology,
Lerner, J. A. (2014). Making democracy fun: How game design Education Connections (the TEC Series). New York, NY:
can empower citizens and transform politics. Cambridge, Teachers College Press.
MA: MIT Press. Sweller, J. (1994). Cognitive load theory, learning difculty,
Logie, R., Baddeley, A., Donchin, E., & Sheptak, R. (1989). and instructional design. Learning and Instruction, 4,
Working memory in the acquisition of complex cognitive 295312. https://doi.org/10.1016/0959-4752(94)90003-5.
skills. Acta Psychologica, 71, 5387. https://doi.org/ Unsworth, N., Heitz, R. P., Schrock, J. C., & Engle, R. W.
10.1016/0001-6918(89)90005-X. (2005). An automated version of the operation span task.
Lucas, K., & Sherry, J. L. (2004). Sex differences in video Behavior Research Methods, 37, 498505. https://doi.org/
game play: A communication-based explanation. 10.3758/BF03192720.
Communication Research, 31, 499523. Van Merrienboer, J. J., & Sweller, J. (2005). Cognitive load
Mayer, R. E., & Johnson, C. I. (2010). Adding instructional theory and complex learning: Recent developments and
features that promote learning in a game-like environment. future directions. Educational Psychology Review, 17,
Journal of Educational Computing Research, 42, 241265. 147177. https://doi.org/10.1007/s10648-005-3951-0.
Mayer, R. E., & Moreno, R. (2003). Nine ways to reduce cognitive Vorderer, P., Knobloch, S., & Schramm, H. (2001). Does
load in multimedia learning. Educational Psychologist, 38, entertainment suffer from interactivity? The impact of
4352. https://doi.org/10.1207/S15326985EP3801_6. watching an interactive TV movie on viewers experience
McNamara, D. S., Kintsch, E., Songer, N. B., & Kintsch, W. of entertainment. Media Psychology, 3, 343363. https://
(1996). Are good texts always better? Interactions of text doi.org/10.1207/S1532785XMEP0304_03.
coherence, background knowledge, and levels of Wise, K., Bolls, P. D., & Schaefer, S. R. (2008). Choosing and
understanding in learning from text. Cognition and Instruction, reading online news: How available choice affects cognitive
14, 143. https://doi.org/10.1207/s1532690xci1401_1. processing. Journal of Broadcasting & Electronic Media, 52,
Moreno, R., & Mayer, R. (2007). Interactive multimodal 6985. https://doi.org/10.1080/08838150701820858.
learning environments. Educational Psychology Review, Yuan, K., Steedle, J., Shavelson, R., Alonzo, A., & Oppezzo,
19, 309326. https://doi.org/10.1007/s10648-007-9047-2. M. (2006). Working memory, uid intelligence, and science
Nakamura, J., & Csikszentmihalyi, M. (2002). The concept of learning. Educational Research Review, 1, 8398. https://
ow. Handbook of positive psychology, 89105. doi.org/10.1016/j.edurev.2006.08.005.

2017 John Wiley & Sons Ltd

View publication stats

Vous aimerez peut-être aussi