734 CYBERPSYCHOLOGY & BEHAVIOR Volume 7, Number 6, 2004 © Mary Ann Liebert, Inc. 1Departamento de Personalidad, Facultad de Psicologia, Universitat de València, València, Spain. 2Universitat Jaume I de Castellón, València, Spain. 3Universitat Politècnica de València, València, Spain. Immersion and Emotion: Their Impact on the Sense of Presence R.M. BAÑOS, Ph.D.,1 C. BOTELLA, Ph.D.,2 M. ALCAÑIZ, Ph.D.,3 V. LIAÑO, B.A.,2 B. GUERRERO, Ph.D.,2 and B. REY, M.S.3 ABSTRACT The present study is designed to test the role of immersion and media content in the sense of presence. Specifically, we are interested in the affective valence of the virtual environments. This paper describes an experiment that compares three immersive systems (a PC monitor, a rear projected video wall, and a head-mounted display) and two virtual environments, one in- volving emotional content and the other not. The purpose of the experiment was to test the in- teractive role of these two media characteristics (form and content). Scores on two self-report presence measurements were compared among six groups of 10 people each. The results sug- gest that both immersion and affective content have an impact on presence. However, immer- sion was more relevant for non-emotional environments than for emotional ones. INTRODUCTION THE SENSE OF PRESENCE has usually been consid-ered the key of virtual reality (VR). Although there is not a common definition of presence, there is a consensus to define it as a multi-component construct. Most authors1–6 agree that presence is de- termined by two general categories of variables: media characteristics and user characteristics. Media characteristics are divided into media form and media content variables. Media form includes the properties of a display medium (e.g., the extent of sensory information presented, the degrees of con- trol that users have over positioning their sensors within the environment, users’ ability to modify as- pects of the environment). Media content includes the objects, actors and events represented by the medium. Finally, user characteristics refer to rele- vant individual aspects ranging from age, gender or cultural variables to users’ perceptual, cognitive, motor abilities, prior experience with mediated ex- periences, willingness to suspend disbelief, and personality differences. Media form characteristics have a significant impact on the sense of presence. Nobody doubts the importance of immersion, interaction, and per- ceptual realism. However, presence research has overemphasized these factors, and sometimes they have been used erroneously to describe the experi- ence of presence. As Schubert et al.7 point out, in some theoretical models, the sense of presence has been seen as the outcome, or a direct function of immersion. Therefore, it has been assumed that the more inclusive, extensive, surrounding, and vivid the virtual environment (VE), the higher the sense of presence.7 There have been some attempts.

1. 734
CYBERPSYCHOLOGY & BEHAVIOR
Volume 7, Number 6, 2004
© Mary Ann Liebert, Inc.
1Departamento de Personalidad, Facultad de Psicologia,
Universitat de València, València, Spain.
2Universitat Jaume I de Castellón, València, Spain.
3Universitat Politècnica de València, València, Spain.
Immersion and Emotion:
Their Impact on the Sense of Presence
R.M. BAÑOS, Ph.D.,1 C. BOTELLA, Ph.D.,2 M. ALCAÑIZ,
Ph.D.,3 V. LIAÑO, B.A.,2
B. GUERRERO, Ph.D.,2 and B. REY, M.S.3
ABSTRACT
The present study is designed to test the role of immersion and
media content in the sense of
presence. Specifically, we are interested in the affective valence
of the virtual environments.
This paper describes an experiment that compares three
immersive systems (a PC monitor, a
rear projected video wall, and a head-mounted display) and two
virtual environments, one in-
volving emotional content and the other not. The purpose of the
experiment was to test the in-
teractive role of these two media characteristics (form and
content). Scores on two self-report

2. presence measurements were compared among six groups of 10
people each. The results sug-
gest that both immersion and affective content have an impact
on presence. However, immer-
sion was more relevant for non-emotional environments than for
emotional ones.
INTRODUCTION
THE SENSE OF PRESENCE has usually been consid-ered the
key of virtual reality (VR). Although
there is not a common definition of presence, there
is a consensus to define it as a multi-component
construct. Most authors1–6 agree that presence is de-
termined by two general categories of variables:
media characteristics and user characteristics. Media
characteristics are divided into media form and
media content variables. Media form includes the
properties of a display medium (e.g., the extent of
sensory information presented, the degrees of con-
trol that users have over positioning their sensors
within the environment, users’ ability to modify as-
pects of the environment). Media content includes
the objects, actors and events represented by the
medium. Finally, user characteristics refer to rele-
vant individual aspects ranging from age, gender
or cultural variables to users’ perceptual, cognitive,
motor abilities, prior experience with mediated ex-
periences, willingness to suspend disbelief, and
personality differences.
Media form characteristics have a significant
impact on the sense of presence. Nobody doubts
the importance of immersion, interaction, and per-
ceptual realism. However, presence research has
overemphasized these factors, and sometimes they

3. have been used erroneously to describe the experi-
ence of presence. As Schubert et al.7 point out, in
some theoretical models, the sense of presence has
been seen as the outcome, or a direct function of
immersion. Therefore, it has been assumed that the
more inclusive, extensive, surrounding, and vivid
the virtual environment (VE), the higher the sense
of presence.7
There have been some attempts to distinguish
presence from immersion. Slater8,9 defined immer-
sion as an objective description of the technology,
while the sense of presence is a subjective experi-
ence and only quantifiable by the user experiencing
it. Likewise, Kalawsky4 states that presence is es-
sentially a cognitive or perceptual parameter, whilst
immersion essentially refers to the physical extent
of the sensory information and is a function of the
enabling technology.
13865C15.PGS 1/7/05 1:13 PM Page 734
However, it would be misleading to assume a
one-to-one relationship between immersion and
presence.7 The other characteristics (media content
and user characteristics) must be also taken into ac-
count. Some VR studies have proved that users can
feel present even in the impoverished environ-
ments world that some VR currently provides.10
For example, Pausch et al.11 analyzed a sample of
Walt Disney World’s Epcot, and their results
showed that the aspects of immersive interfaces
(displays, graphics, and control device quality)
were important to the users; significantly, however,

4. the background stories and goals as well as the
“physics fidelity” (e.g., motion) of the rides had
even more impact on the users experience.
The present study is addressed to test the role of
immersion and media content in the sense of pres-
ence. Specifically, we are interested in the affective
valence of the virtual environments. Huang and
Alessi,12 point out that emotions are an essential
part of how people experience the world, and their
study could have important implications for a bet-
ter understanding of the virtual experience. These
authors stated that definitions of presence have
mostly been cognitively or environmentally based,
generally ignoring the emotional aspects of pres-
ence. However, emotions play an important role
in our subjective judgments and automatic re-
sponses, influencing our learning as well as how
we understand, describe and react to the world and
ourselves. In two studies,13,14 we found important
differences in the responses to VR environments
between non-patients and (mental health) patients
that proved the importance of emotions for clinical
users. Emotions may play a role both as determi-
nants and consequences of presence.
Factorial studies using self-report measures of
presence have also shown the importance of emo-
tional engagement. In particular, studies by two dif-
ferent teams reveal very similar factor structures.
Lessiter et al.,5 using ITC-Sense of Presence Inven-
tory (ITC-SOPI) questionnaire, reported a four-factor
solution for presence: physical space, engagement,
naturalness, and a fourth attenuating factor, nega-
tive effects. Schubert et al.,7 using their self-report
presence survey, arrived at a three-factor solution

5. for the presence construct almost identical to the
ones identified by Lessiter et al.: spatial presence,
involvement, and realness. According to these fac-
torial structures, it seems that presence is not only
related to a sense of a physical, spatial environment
(the sense of “being there”), but also to a personal
evaluation of the appeal, and the naturalness/be-
lievability, of both the displayed environment and
its content.5 As Lessiter et al.5 point out, these di-
mensions may contribute to the sense of presence
in an additive way, or in a more complex, interac-
tive manner.
The goal of the present study is not to test
whether people have similar emotional responses
in the virtual environment and in similar circum-
stances in the real world. The goal is to test whether
presence can be enhanced in less immersive virtual
environments by using emotional content. Three
immersive systems will be compared: a head-
mounted display (HMD), a rear projected video
wall, and a PC monitor. According to Kalawsky,4 if
the display presents a full 360º information space,
then it is a “fully immersive system”; if the extent
of the display is less than 360º, it is a “semi-immer-
sive system.” The term “non-immersive system” is
usually reserved for desktop VR systems. This
study compares the sense of presence on these
three immersive systems between two different vir-
tual environments; one involves emotional content
and the other does not. The specific questions ad-
dressed are as follows: (1) Does an emotional vir-
tual environment elicit a higher “subjective” sense
of presence than a non-emotional virtual environ-
ment? (2) If so, does this depend on the immersive

6. characteristics of the system?
MATERIALS AND METHODS
Experimental design
In order to study the role of immersion and affec-
tive content on the subjective sense of presence, the
following variables will be manipulated:
• Immersion: three conditions were considered: a
HMD, a semi-immersive system video wall, and
a PC monitor.
• Affective content: Two virtual environments
were designed, one to induce sadness (emotional
condition) and another in which no mood changes
were expected (neutral condition).
A 2 ! 3 between groups design was used, with six
experimental conditions.
Participants
Sixty participants were recruited for the study
from the Polytechnic University of Valencia, Uni-
versity of Valencia and University Jaume I of
Castellon. There were 37 females and 23 males. The
mean age was 24.78 (SD = 5.847), with a range be-
tween 18 and 49. Groups of 10 participants were
IMMERSION AND EMOTION 735
13865C15.PGS 1/7/05 1:13 PM Page 735

7. randomly allocated to one of six experimental con-
ditions. All participants fulfilled the following in-
clusion criteria: (a) non-history of neurological
disease, head injury, learning disability or mental
disorders; (b) non-history of psychological disor-
ders; (c) non-use of any medication for psychologi-
cal or emotional problems; and (d) scoring lower
than 18 in BDI (Beck Inventory Depression).15
Measures
Beck Depression Inventory. The BDI15 is the most
widely used self-report instrument for measuring
depressive symptom severity in both research
and clinical settings. It is a 21-item self-report
questionnaire. Scores less than 18 are considered
normative.
ITC–Sense of Presence Inventory (ITC-SOPI). The
ITC-SOPI is a post-test subjective presence measure
composed of 44 items, divided in two parts.5 Part A
(6 items) refers to a respondent’s impressions/feel-
ings after a media experience has finished. Part B
(38 items) refers to a respondent’s impressions/
feelings during a media experience. A 1–5-point
Likert scale (from Strongly Disagree to Strongly
Agree) is used for responding to both parts. Factor
analysis showed that this questionnaire measures
four dimensions: Physical space, Engagement, Eco-
logical Validity, and Negative effects. ITC-SOPI:
internal reliability coefficients (alpha) were com-
puted for each of the four factors. Alphas were
high, ranging from 0.94 (Physical Space) to 0.76
(Naturalness).

8. Reality judgment and presence questionnaire (RJPQ).
RJPQ13 is a post-test subjective presence and reality
judgment measure. A short version of this ques-
tionnaire, with 29 items, was used. A 1–10 Likert
scale was used for responding to all items. The fol-
lowing factors were considered: “Quality/Realism”
(11 items regarding the quality and congruence of
the images and sounds, and the influence of quality
on the sense of presence and realness); “Reality
Judgment” (4 items related to the realness of the en-
vironments, the objects and the experience); “Pres-
ence: Positive” (8 items related to the sense of being
in the virtual environment); “Presence: Negative”
(3 items related to the difficulties of feeling pres-
ence); “Interaction/Navigation” (7 items related to
movements and interactions); “Emotional engage-
ment” (3 items related to the emotions felt in the
virtual environment); and “Emotional Indiffer-
ence” (3 items related to boring and disappointing
feelings). The alpha reliability (internal consistency)
for the questionnaire was 0.82.
Virtual environments
A Mood Induction Procedure (MIP) using VR
was used. MIPs are experimental procedures
whose aim is to provoke a transitory emotional
state in an individual in a non natural situation and
in a controlled manner. The mood induced should
be specific and ideally is an experimental analogue
of the mood that would occur in a natural situa-
tion.16 MIPs include a broad diversity of methods
and have proven to be effective in achieving
changes in the target mood.17–19

9. The VR-MIP consists of a neutral environment
(Fig. 1) that progressively changes depending on
the mood state to be evoked in the user. The sce-
nario is a park, that is, a natural and urban ambient
that can be found easily in any city or culture in the
real world. We chose this environment because it
includes elements of nature (trees, flowers, water,
etc.), and because changing some of the light para-
meters (tone, direction, brightness) easily modifies
the aspect of these elements, inducing different
moods in the user. For example, in the case of sad-
ness, the park is grey, it is a cloudy day, the trees
have no leaves, there are no people in the park and
the music that is heard is very sad. In Figure 2,
some views of the “sad park” are showed.
For the present study, two variations of this VR-
MIP were used. The goal of one was to evoke sad-
ness (emotional condition), and the goal of the
other one was to not evoke any specific mood (neu-
tral condition). In order to build the different envi-
ronments, variations of every one of following
elements were included: music, narratives, Velten
736 BAÑOS ET AL.
FIG. 1. One view of the neutral park.
13865C15.PGS 1/7/05 1:13 PM Page 736
self-statements*,20 as well as pictures (selected from
International Affective Picture System IAPS21),
movies, and autobiographical recalls.

10. The environment content is as follows: Users
listen to a short history corresponding to the emo-
tional experimental condition (sadness versus neu-
tral). A woman’s voice guides users through a virtual
walk. From the beginning, a piece of music is heard
(emotional condition: “Adagio for Strings-Choral”
by Samuel Barber; neutral condition: “Nothing
Spectacular” by Michael Lindh, which was com-
posed by Michael Lindh from Interactive Institute
and was validated as a neutral MIP in the EMMA
project.). The initial appearance of the environment
is the same for all users. However, the aspect
changes shortly thereafter, depending on the in-
tended emotional condition. Users have two min-
utes to freely explore the park.
Then, they are asked to go to the center of the
park, where a bandstand is located. On five of the
sides of the stand (it is an eight-faced polyhedron),
a statement of the Velten20 technique appears in a
disordered manner and users must order it. The
content of the statements depends on the emotional
condition (Table 1). For each sentence, users have to
choose a picture from four options, the one that
best represents (according to them) the meaning of
the sentence (selected from IAPS21). Users are
asked to get involved in the contents of each sen-
tence for 45 seconds, and to think about the per-
sonal meaning of each statement. After that, they
can walk around the virtual park again for two
minutes. Then, users are asked to go to the cinema
to watch a short film (scenes from “The Champ”;
emotional condition: scenes from “A True Story”
for neutral). Once the cinema session is finished,
users are asked to produce an autobiographical re-

11. call in a loud voice, similar to the experiences they
encountered in the park.
Hardware
The workstations for running the virtual envi-
ronments were PC based computers with high-end
graphics capability, with 128 Mb of memory for
graphics and textures. Regarding the interaction
device, a joystick was used. This device was config-
ured to have different modes of use, so that press-
ing a button alternated between the navigation and
interaction modes. The display’s devices included
the following:
• PC monitor: 17-inch monitor with a resolution of
1024 ! 768 pixels.
• HMD: A HMD (model 800 from Fifth Dimension
Corporation, Irvine, CA) with a head-tracking
device (model intertrax2 from Intersense, Bed-
ford, MA)
• Big screen: A rear projected video wall setup was
created using a metacrilate retro-projected screen
of 400 ! 150 cm. The retro-projection option al-
lowed users to walk near the screen without
blocking the image or projecting shadows on the
screen. Resolution projectors were 1024 ! 768
pixels with a power of 2000 lumens; however, it
was limited to a power of 1000 lumens in order
to make users feel more comfortable.
Procedure
Participants were given the following description

12. of the study: “This is an experiment about virtual
reality. First of all, I am going to ask you a few ques-
IMMERSION AND EMOTION 737
*This is a MIP developed by Velten,20 wherein mood induction
is achieved by means of statements written in first person,
relative to
the mood. Subjects are asked to read the statements, and to try
to feel a mood similar to the one described in them.
FIG. 2. One view of the sad park.
TABLE 1. VELTEN SELF-STATEMENTS
Neutral sentences Sad sentences
Japan is a set of Life seems sad and
islands. senseless to me.
The house is for sale. I make people unhappy.
The train travels from I fail in everything.
Madrid to Sevilla.
The ship was ancient. I have no future.
The doorkeeper was I am worthless.
dressed in red.
13865C15.PGS 1/7/05 1:13 PM Page 737
tions. After that you will practice in a training vir-
tual environment. During this time I can help you
if you have any doubts. Later, you will stay alone

13. in another virtual environment. When virtual ex-
perience finishes you have to fill in some ques-
tionnaires.” Participants were provided informed
consent to take part in the study, and were asked to
complete a short screening interview and BDI, in
order to ask about exclusion criteria. Then they were
randomly assigned to one of six experimental condi-
tions, and practiced in the training virtual environ-
ment. After the VR-MIP, participants completed the
two presence questionnaires. All participants were
debriefed following the experiment.
RESULTS
Means and standard deviation for the question-
naires can be found in Table 2. Analysis of variance
was conducted on the presence measures, with
emotional conditions (sadness versus neutral) and
immersive conditions (Monitor, Big screen, and
HMD) as between-groups factors. The dependent
variables were the various factors of presence ques-
tionnaire measures (ITC-SOPI and JRPQ).
Regarding ITC-SOPI, a main effect of “emotional
condition” was found for engagement (F(1.53)=
3.99, p < 0.05), and ecological validity (F(1.53) =
3.98, p < 0.05). A main effect of “immersive condi-
tion” was only found for negative effects (F(2.53) =
6.06, p < 0.004). An interaction effect emotional !
immersive conditions was found for engagement
(F(2.53) = 3.59, p < 0.03) and ecological validity
(F(2.53) = 3.12, p < 0.05). No other significant effects
were found. In general, the sad group scored
higher in engagement and ecological validity than
the neutral group. The HMD condition provoked

14. more negative effects than the other two immersive
conditions. Finally, with respect to interaction ef-
fects, post-hoc analysis revealed that monitor con-
ditions produced a different pattern in sad and
neutral conditions. The sad group using a monitor
scored higher on engagement and ecological valid-
ity than the neutral group.
Regarding JRPQ, a main effect of “emotional
condition” was found for reality judgment (F(1.54)
= 3.77, p < 0.05), emotional engagement (F(1.54) =
20.15, p < 0.000), and emotional indifference (F =
(1.54) = 8.44, p < 0.005). Quality/realism (F(1.54) =
3.6, p < 0.06) almost reached statistical significance.
In general, sad groups scored higher on reality
judgment, emotional engagement and quality/re-
alism than neutral groups, while neutral groups
scored higher on emotional indifference. A main
effect of “immersive condition” was found for qual-
ity/ realism (F(2.54) = 4.85, p < 0.012); and interac-
tion/navigation (F = (2.54) = 4.41, p < 0.017). Big
screen groups scored higher on quality/realism
and interaction/navigation. No other significant
effects were found. However, an interaction effect
emotional ! immersive conditions was almost sta-
tistically significant for reality judgment (F(2.54) =
2.95, p < 0.06); and presence positive (F(2.54) = 2.91,
p < 0.06). Post-hoc tests revealed that there were no
differences between sad and neutral conditions
738 BAÑOS ET AL.
TABLE 2. MEANS AND STANDARD DEVIATIONS
Monitor Big Screen HMD

15. Sad Neutral Sad Neutral Sad Neutral
Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD
ITC-SOPI
Physical space 3.34 0.72 2.68 0.66 3.09 0.57 3.08 1.27 2.85
0.39 2.96 0.68
Engagement 3.45 0.55 2.70 0.71 3.55 0.46 3.22 0.53 3.01 0.43
3.17 0.52
Ecological validity 3.64 0.64 2.60 0.67 3.16 0.81 3.00 0.85 3.21
0.56 3.12 0.84
Negative effects 1.78 0.85 1.68 0.69 1.72 0.62 1.67 0.93 2.41
0.81 2.60 0.94
JRPQ
Quality/realism 7.46 1.68 6.36 1.52 7.58 0.53 7.18 1.14 6.30
1.47 5.74 1.70
Reality judgment 6.53 1.62 3.85 1.06 5.70 1.21 5.63 2.76 5.48
2.29 5.33 1.05
Presence positive 6.71 1.74 4.51 1.31 6.04 1.25 6.05 1.63 5.11
1.64 5.16 1.36
Presence negative 3.47 1.74 4.40 1.96 4.27 1.63 3.67 2.19 5.17
1.27 4.40 1.81
Interaction/navigation 7.09 1.38 6.44 1.43 7.54 0.47 7.09 1.60
6.23 1.15 5.86 1.74
Emotional engagement 8.30 1.93 5.30 1.65 8.30 1.18 5.37 2.33
7.37 2.76 5.30 1.56
Emotional indifference 2.60 2.22 4.30 1.22 1.83 1.22 3.93 1.40
3.00 2.71 3.93 1.42
13865C15.PGS 1/7/05 1:13 PM Page 738
when participants were immersed using a HMD.
When they were immersed using the big screen,

16. there were differences between sad and neutral
conditions for emotional engagement and emo-
tional indifference (sad group scoring higher on
emotional engagement and lower on emotional in-
difference). However, there were many differences
when participants were immersed using only a
monitor. The sad group using this less immersive
system scored higher than the neutral group on re-
ality judgment, presence positive, and emotional
engagement, and lower on emotional indifference.
DISCUSSION
Results of this study show that both affective
content and immersion have an important effect on
the sense of presence. First, regarding immersion,
ITC-SOPI results revealed that the only difference
among the three immersive conditions was in the
category of “negative effects.” Not surprisingly, the
HMD provoked more undesirable effects (dizzi-
ness, disorientation, nausea) than the other immer-
sive systems. However, this questionnaire did not
reveal other significant differences among the three
immersive systems. RJPQ data revealed that the big
screen elicited higher “Quality/realism” and “Nav-
igation/interaction” subjective scores than the other
two conditions. Although HMD is considered a
fully immersive system, because it displays a 360º
information space, the size of the screen seems to
be more relevant in order to achieve realism. Nev-
ertheless, it is interesting to note that big screen
users rated higher on “navigation/interaction,” al-
though the same navigation device (a joystick) was
used by all participants. However, HMD users
were also provided with a head tracking device (In-
tertrax II). It might be that the negative effects of

17. this device made the navigation more difficult.
Regarding affective content, results show differ-
ences between emotional and neutral environ-
ments in presence measurements. Both ITC-SOPI
and RJPQ results show that the emotional environ-
ment seems to be more engaging, natural, believ-
able and real to users than the neutral environment.
We believe that the most interesting results of
this study are those of interaction effects among af-
fective content and immersive systems. According
to data, the sense of presence in the non emotional
environment depends mainly on immersion. Both a
HMD and a big screen elicited a higher sense of
presence than a non-immersive system, namely, a
PC monitor. However, in an emotional environ-
ment a PC monitor was able to elicit a high sense of
presence, in the same way that a big screen was. On
the other hand, the HMD condition was not the
most presence-enhancing technology; as stated
previously, these data could be explained by the
less comfortable HMD setup.
Our results illustrate that presence is not a direct
function of immersion alone. It is misleading to as-
sume a one-to-one relationship between immersion
and presence.7 As Ijjselsteijn22 states, although the
breadth and depth of sensory experience is impor-
tant in improving the media experience, intensity
does not equal quality. “The basic appeal of media
still lies in its content, the storyline, the ideas and
emotions that are being communicated.”22 This
does not mean to ignore the media form altogether;
“the psychological impact of content, both good

18. and bad, exciting and boring, depends to a large ex-
tent on the form in which it is represented.”22
Nevertheless, our results indicate that efforts
must not be solely focused on technology. As
Heeter23 points out, “presence research has empha-
sized engineering the senses more strongly that it
has engineering the mind . . . Sensory realism is
certainly an important influence on presence, but
there is more to the story”. It is important to re-
member, as Biocca24 does, that Munsterberg, the
first psychologist to study media in 1916, hinted at
an issue which sometimes have been not taken into
account: media obey laws of the mind. Presence is a
user experience and it is not intrinsically bound to
any specific type of technology, but is rather a
product of the mind.2 Biocca24 also reminds us of
the celebrated phrase of Bricken, from the 1990
SIGGRAPH conference: “Psychology is the physics
of virtual reality.” According to Biocca, this sen-
tence implies that, like physics, psychology holds a
key to our understanding of reality. Therefore, VR
“has less to do with simulating physical reality per
se; rather it simulates how the mind ‘perceives’
physical reality.”24 Therefore, presence research
will have to extend beyond a search for realism or
fidelity only.2
Our study has been focused in only one of the
media form characteristics, immersion, but there
are other media form variables that are also very
relevant, one of which is interaction. According to
Ijsselsteijn,25 interactivity appears to be a more im-
portant factor than immersion. Interactive, non-
realistic displays are able to engender substantial
levels of presence. Therefore, it would be also im-

19. portant to test a possible interaction effect between
interactivity and affective content of virtual envi-
ronments. “Being there” has been considered the
ability to “do there,”25 but this study also adds the
possibility of “feeling there.” Thus it can be said, “I
IMMERSION AND EMOTION 739
13865C15.PGS 1/7/05 1:13 PM Page 739
can do here, therefore I am here,” but also, “I feel
here, therefore I am here.”
It is important, from both theoretical and applied
perspectives, to determine the most critical ele-
ments in feeling presence for different VR applica-
tions. One of these VR applications is psychological
treatments. Results in this area have been promis-
ing and presence research must contribute to
answering important questions such as: What ele-
ments are fundamental in order to achieve the
sense of presence in therapeutic applications? Our
results indicate that if the focus is on eliciting emo-
tions with the goal of reducing or modifying them,
immersion factors could be less important than a
carefully content design. This content design would
have to include those elements relevant for every
specific emotional problem, that is, those elements
with the potential of activating emotions. There-
fore, the focus must be on the psychological aspects
more than on the technical aspects.
ACKNOWLEDGMENTS

20. This research was funded through the EMMA pro-
ject (IST-2001–39192), funded by the European Com-
munity: V Framework Programme (IST Programme,
8th call, “Future and Emerging Technologies.”
REFERENCES
1. Barfield, W., Zelter, D., Sheridan, T.B., et al. (1995).
Presence and performance within virtual environ-
ments. In: Barfield, W., & Furness, T.A. (eds.), Virtual
environments and advanced interface design. Oxford:
Oxford University Press.
2. IJsselsteijn, W.A., & Riva, G. (2003). Being there: the
experience of presence in mediated environments.
In: Riva, G., Davide, F., & Ijsselsteijn, W. (eds.), Being
there: concepts, effects and measurement of user presence
in synthetic environments. Amsterdam: Ios Press.
3. IJsselsteijn, W.A., de Ridder, H., Freeman, H., et al.
(2000). Presence: concept, determinants and mea-
surement. Proceedings of the SPIE 2000 3959: 520–529.
4. Kalawsky, R.S. (2000). The validity of presence as a
reliable human performance metric in immersive en-
vironments. Presented at Presence 2000: Interna-
tional Workshop on Presence, Delft, Netherlands.
5. Lessiter, J., Freeman, J., Keogh, E., et al. (2001). A
Cross-Media Presence Questionnaire: the ITC Sense
of Presence Inventory. Presence: Teleoperators and Vir-
tual Environments 10:282–297.
6. Lombard, M., & Ditton, T. (1997). At the heart of it
all: the concept of presence. Journal of Computer Com-
munication 3:2.

21. 7. Schubert, T.W., Friedmann, F., & Regenbrecht, H.T.
(2001). The experience of presence: factor analytic in-
sights. Presence: Teleoperators & Virtual Environments
10:266–281.
8. Slater, M. (1999). Measuring presence: a response to
the Witmer and Singer presence questionnaire. Pres-
ence: Teleoperators and Virtual Environments 8: 560–565.
9. Slater, M., Usoh, M., & Steed, A. (1994). Depth of
presence in virtual environments. Presence: Teleopera-
tors and Virtual Environments 3:130–144.
10. Riva, G., Loreti, P., Lunghi, M. et al. (2003). Presence
2010: the emergence of ambient intelligence. In: Riva,
G., Davide, F., & Ijsselsteijn, W. (eds.), Being there:
concepts, effects and measurement of user presence in syn-
thetic environments. Amsterdam: Ios Press.
11. Pausch, R., Snoody, J., Taylor, R., et al. (1996). Dis-
ney’s Alladin. First steps toward storytelling in vir-
tual reality. Presented at the 23rd Annual Conference
on Computer Graphics, Association for Computing
Machinery, Inc., New York.
12. Huang, M.P., & Alessi, N.E. (1999). Mental health im-
plications for presence. CyberPsychology and Behavior
2:15–18.
13. Baños, R.M., Botella, C., García-Palacios, A., et al.
(2000). Presence and reality judgement in virtual en-
vironments: A unitary construct? Cyberpsychology and
Behavior 3:327–335.
14. Baños, R.M., Botella, C., García-Palacios, A., et al.

22. (2001). The role of reality judgment and presence in
virtual environments for clinical psychology. Pre-
sented at the World Congress of the Behavioural and
Cognitives Therapies, Vancouver, Canada.
15. Beck, A.T., Ward, C.H., Mendelson, M., et al. (1961).
An inventory for measuring depression. Archives of
General Psychiatry 4:561–571.
16. García-Palacios, A., Baños, R. (1999). Eficacia de dos
procedimientos de inducción del estado de ánimo e
influencia de variables moduladoras. Revista de Psi-
copatología y Psicología Clínica 4:15–26.
17. Gerrards-Hesse, A., Spies, K., & Hesse, F.W. (1994).
Experimental inductions of emotional states and
their effectiveness: a review. British Journal of Psychol-
ogy 85:55–78.
18. Martin, M. (1990). On the induction of mood. Clinical
Psychology Review 10:669–697.
19. Westermann, R., Spies, K., Stahl, G., et al. (1996). Rel-
ative effectiveness and validity of mood induction
procedures: a meta-analysis. European Journal of So-
cial Psychology 26:557–580.
20. Velten, E. (1968). A laboratory task for induction of
mood states. Behaviour Research and Therapy 6:473– 482.
21. Lang, P.J., Bradley, M.M., & Cuthbert, B.N. (1995). In-
ternational Affective Picture System (IAPS): Technical
Manual and Affective Ratings. NIMH Cent. Study
Emot. Atten. University of Florida.
22. IJsslesteijn, W.A. (2003). Presence in the past: what

23. can we learn from media history? In: Riva, G., Da-
vide, F., & Ijsselsteijn, W. (eds.), Being there: concepts,
effects and measurement of user presence in synthetic en-
vironments. Amsterdam: Ios Press.
740 BAÑOS ET AL.
13865C15.PGS 1/7/05 1:13 PM Page 740
23. Heeter, C. (2003). Reflections on real presence by a
virtual person. Presence: Teleoperators and Virtual En-
vironments 12:335–345
24. Biocca, F. (2003). Preface. In: Riva, G., Davide, F., &
Ijsselsteijn, W. (eds.), Being there: concepts, effects and
measurement of user presence in synthetic environments.
Amsterdam: Ios Press.
25. IJsselsteijn, W.A. (2002). Elements of a multi-level the-
ory of presence: phenomenology, mental processing
and neural correlates. Presented at Presence 2002: In-
ternational Workshop on Presence, Porto, Portugal.
Address reprint requests to:
Rosa M. Baños, Ph.D.
Departamento de Personalidad
Facultad de Psicologia
Universitat de València
Avda. Blasco Ibañez, 21
46010-València, Spain
E-mail: [email protected]

24. IMMERSION AND EMOTION 741
13865C15.PGS 1/7/05 1:13 PM Page 741
Stress training and simulator complexity: why sometimes more
is less
Jennifer G. Tichon* and Guy M. Wallis
The School of Human Movement Studies, The University of
Queensland, St Lucia, 4072, Australia
(Received 20 January 2009; final version received 15 October
2009)
Through repeated practice under conditions similar to those in
real-world settings, simulator training prepares an
individual to maintain e!ective performance under stressful
work conditions. Interfaces o!ering high fidelity and
immersion can more closely reproduce real-world experiences
and are generally believed to result in better learning
outcomes. However, absolute fidelity in stress training is not
critical for skills to be transferable. The present study
compared the performance outcomes achieved by trainees using
two di!erent simulator types to complete a training
program aimed at improving decision-making skills. The
purpose of this research was to assess both the overall level
of training e!ectiveness and to determine whether performance
levels were influenced when high (160 degree curved
wide screen) versus low fidelity (small cab-based flat screen)
simulator types were in use. Sixty-three train drivers

25. drove for 40 min on a simulated track on which they
encountered four major high stress driving events. One year
later, 42 of the original drivers returned and repeated the
training scenario a second and third time. Results revealed
trainees using the lower fidelity flat screen simulator made
fewer errors in both years than trainees using the high
fidelity curved screen simulator. The implications of these
results are discussed.
Keywords: task analysis; simulator; stress; cognitive training;
presence
1. Introduction
In work environments where heavy, human operated
machinery is at work, accidents are an unfortunate but
often inevitable part of everyday life. An essential
element, in training to respond to workplace accidents
e!ectively, is how to train sta! to cope with ill-defined
problems in the face of high stress environments
constrained by factors such as insu"cient or unreliable
information, time constraints and or conditions of
threat of physical injury (Salas et al. 1999). Emergency
response skills, particularly those that degrade under
stress such as critical thinking and decision-making,
have been highlighted across a number of industries as
in need of attention.
The question of how to most e!ectively train
higher-order mental abilities in these situations is now
emerging as a focus of research interest. Suitable
training in critical thinking under stress can hugely
reduce the impact of workplace accidents, not only in
terms of the immediate physical well-being of sta! and
clients, but also the long-term mental health of those
involved. Simulators through recreating real-world

26. situations in virtual environments provide a compel-
ling opportunity for meeting the need for high a!ect
training. Accordingly simulator training has been
established as a core component of safety training
programs. High fidelity, interactive simulation is
typically achieved through complete immersion in
brief, stressful and complex scenarios.
2. Stress exposure training via simulation
When training goes beyond the acquisition of stan-
dard, required knowledge and abilities and instead
aims to prepare trainees to perform e!ectively in a
stressful environment, it is commonly referred to as
stress exposure training (Driskell and Johnston 1998).
Stress exposure training focuses on developing those
cognitive skills required to maintain e!ective perfor-
mance under stress. The overall goal of training via
simulation is to build confidence in sta! in their own
ability to perform under adverse conditions (Stetz et al.
2006). Sub-goals of stress exposure training include
gaining specific knowledge of and familiarity with the
operational environment to assist trainees to form
accurate expectations of the environment, increasing
their ability to predict outcomes, avoid errors and
decrease their propensity to be distracted by novel
sensations (Driskell and Johnston 1998, Hulse and
Memon 2006). Simulator training has been shown to
be e!ective when the trainee experiences success or a
sense of mastery during the training (Maschuw et al.
2008). This confidence translates into being less
aroused physiologically, less distracted and more likely
to focus on the task.
*Corresponding author. Email: [email protected]

27. Behaviour & Information Technology
Vol. 29, No. 5, September–October 2010, 459–466
ISSN 0144-929X print/ISSN 1362-3001 online
! 2010 Taylor & Francis
DOI: 10.1080/01449290903420184
http://www.informaworld.com
Simulators are used for instruction and training in
areas such as commercial and military aviation, battle
field management, building construction, and first
responder agencies because they provide a safe
alternative to replicating high-risk real-world training
scenarios. Real-world renditions of these situations are
too costly and dangerous to provide opportunities for
sta! to repeatedly practise responding to high stress
workplace situations (Tichon et al. 2003). It is the
ability of immersive simulation to engender the same
psychological and psychological responses experienced
in the real world that has been seen as valuable
(Romano and Brna 2001).
3. High fidelity or low fidelity?
A crucial element of the approach is to provide the
opportunity for repeated practice under operational
conditions similar to those likely to be encountered in
real-world settings. The aim is to incite the same
emotional (stress) reaction and stems from research
demonstrating that for some tasks normal training
procedures did not improve task performance when
the task was later performed under stress conditions
(Zakay and Wooler 1984).

28. The use of simulator technology alone, however,
cannot ensure a successful training outcome without
prior consideration of the best use of its features
(Cannon-Bowers and Salas 1998, Wallis et al. 2007).
Prior research has described a number of variables of
interest to simulation-based critical decision-making
training but the best use of these is not always agreed on
(Thomas 2003). There is a debate over the level of
fidelity required for successful training outcomes. Some
suggest that when using simulation for cognitive skills
training, the fidelity should only match the require-
ments of the training such as facilitating reflection and
learning in the training experience (Ho!man et al.
2001). Yet, it is also recognised that the similarity of the
training environment to the actual conditions under
which the trainee will perform in the real world is an
important factor in simulation training design. While
realistic simulations do not always translate directly to
training e!ectiveness, good outcomes have been re-
ported when simulations have reproduced realistic
tasks and a!orded trainees an engrossing experience
they can relate to their real world (Baker et al. 2005). A
study of simulator training in underground cave
structures reported task performance was two to three
times faster in the high-immersion condition, and
question responses were 3 to 10 times more accurate
(Bowman and McMahan 2007).
Despite increasing a!ordability high-end immersive
technologies such as head mounted displays and multi-
screen stereoscopic projection displays remain too
expensive for widespread use. In training applications
significant improvements in performance using simu-
lators in comparison to alternative training support
tools would have to be proven to justify the expense.

29. Consequently, questions such as ‘How much immer-
sion is enough?’ and ‘When are driving environments
real enough?’ are being more and more frequently
asked (Walshe et al. 2005, Bowman and McMahon
2007). Research has suggested that in driving applica-
tions specifically di"culties can arise if users are not
immersed su"ciently in simulated driving scenes
(Nanyue et al. 2006). Is there a compromise between
the traditional alternatives of textbooks and classroom
teaching and the high-end, high fidelity that can still
produce a level of realism su"cient to gain the required
performance outcomes of stress exposure training?
The overall purpose of this research was to assess
the level of training e!ectiveness in terms of enhanced
decision-making under stress. To achieve this, compar-
isons among performance outcomes over three di!erent
training exposures was made. Additionally, the study
design allowed for comparison of the performance
results of trainees using a high fidelity curved wide-
screen simulator with the performance of trainees using
a smaller cab-based flat screen simulator. Reactions to
four separate stressful events were examined for e!ect
of simulator type on learning outcomes.
As it is the illusion of being in a di!erent place that
is essential for the e!ectiveness of these training
applications (Romano and Brna 2001, Baker et al.
2005), the two simulator types were also compared in
terms of their e!ectiveness in generating immersion
and thereby engaging the driver’s attention. Presence is
a variable widely accepted as being positively related to
enhanced learning and performance in immersive
training environments and therefore highly desirable
(Witmer and Singer 1998). It facilitates the environ-
ment replicated in the simulator being interpreted as

30. real by the trainee thereby making it more likely
decision-making skills developed will be transferred to
the real world. Presence has been described ‘as the
subjective experience of being in one place or environ-
ment, even when one is physically situated in another’
(Witmer and Singer 1998). Research investigating the
role of presence in facilitating the e!ectiveness of
simulator-based educational training indicates the two
seem to be inextricably tied.
4. Method
4.1. Participants
4.1.1. Year one
In 2006, the first year of the project, 63 participants
volunteered to take part in the study. The 61 males and
460 J.G. Tichon and G.M. Wallis
2 females possessed an average of 14.8 (SD ! 9.68)
years’ experience driving trains, with a range between 1
and 30 years of experience. The drivers were aged from
29 to 58 years of age with a mean age of 43.3
(SD ! 8.14) years.
Twelve participants were trained in the wide-screen
reality centre while 51 were trained in the cab-based
simulator. The unequal sample size for the simulators
was due to drivers being given the opportunity to self-
select the simulator they would use for training. This
training was part of their annual training and drivers
were allowed to choose their simulator as per normal

31. operating procedures. As drivers were being assessed
via their simulation performance they were allowed to
select the simulator they felt the most comfortable and
confident using. Self-selection was also allowed for the
study as it did not result in a driver being randomly
assigned a simulator with which they were unfamiliar.
A no-control group design was used at the request of
the rail company which required all drivers to undergo
the annual safety training.
4.1.2. Year two
Forty-two participants who participated in the 2006
study above returned in 2007. However, due to a
malfunction with the wide-screen simulator eight of
these drivers were reassigned the cab-based simulator.
As this was not the simulator they had used in 2006
these eight sets of results could not be used in this
study. Because of the expense of bringing active drivers
o!-line for the research it could not be rescheduled for
another time. Of the remaining 34 participants, 33 were
males and 1 was a female. They possessed an average
of 12.9 (SD ! 9.94) years’ experience driving trains,
with a range between 1 and 32 years of experience. The
drivers were aged from 30 to 59 years of age with a
mean age of 43.5 (SD ! 8.66) years. Again partici-
pants volunteered to take part in the study.
4.2. Equipment
Results were obtained from two simulator types:
(1) Cab-based. A reconfigurable, full-size, in-cab
simulator which can reproduce the arrange-
ment of standard in-cab train controls. Visual
display input was provided via a flat projection

32. screen (3 m wide 6 2 m high) and driven by a
single-channel, SGI Onyx 2 (Four 400 MHz
CPUs).
(2) Wide-screen. A curved screen reality centre
7.3 m in diameter o!ering a 160 degree field-of-
view from a centrally located train control
desk. The desk has force-feedback controls
but no enclosure. The system is driven by a
three-channel SGI Onyx 3000 (Twenty-four
500 MHz CPUs).
4.3. Training scenario
The training scenario involved entering a worksite
operating under Track Work Authority (TWA),
encountering detonators and an outer hand signaller
at the bottom of the flyover. Drivers then had to stop
at a station and remember prior warnings when
departing the station before encountering further
detonators, an inner hand signaller and finally the
worksite. A second incident involved a Condition
A!ecting the Network (CAN). Specifically, the CAN
was a failed level crossing attended by hand signallers.
After receiving advice of the failed crossing the driver
next encounters school children trespassing on the
track, followed by a station and finally a tunnel. These
interleaved events were designed to interfere with the
driver’s concentration potentially resulting in the
driver forgetting to prepare for the failed crossing.
Finally, drivers were required to pass a failed absolute
signal at stop immediately before a station located in a
tunnel. Each of these events were further complicated
through increasing the driver’s workload pressures by
rapidly changing operational conditions such as

33. variations to track infrastructure including curvature,
grade, speed and signal siting and sequence. The rate at
which these conditions were changed often resulted in
a driver feeling under intense time pressure. Incidents
also included the receipt of information into the
driving cab via multiple sources, both visual and
auditory.
4.4. Procedures
To determine whether stress training via simulation
was improving performance the decision making skills
of 63 drivers undertaking their annual simulator
training were recorded in 2006. Rail incidents encoun-
tered during the simulator training exposure were
analysed to compile a checklist of correct actions the
driver must perform simultaneously for each key event
in addition to a checklist of possible errors. From the 4
major incidents encountered in the training scenario
outlined above, 36 correct actions were required to
maintain safe and e!ective train travel on the network
and 45 possible errors were identified. All drivers
attended a classroom refresher course on rules and
regulations and undertook a 5 min practice run on the
simulator before their simulation training. Trainers
completed the checklist of correct actions and possible
errors while drivers undertook the 40 min simulator
training run.
Behaviour & Information Technology 461
In 2007, 34 of the drivers who had participated in
2006 returned for training to undertake the same
simulated driving scene using the same simulator type

34. they had been tested on the previous year. This is
referred to as Trial 1. To test the premise that repeated
exposure will inoculate drivers and result in enhanced
decision-making skills, the drivers were also required
to undertake the driving scene after a short half hour
time delay. This is referred to as Trial 2. Comparing
across exposures, Trial 1 against Trial 2, decision
making performance could then be assessed after a
short time delay (on the same day) and after a long
time delay (12 month interval). All the same perfor-
mance data were gathered during the first and second
sessions as was collected in 2006 and drivers used the
same simulator type which they had used in the first
instance.
In 2007 in between the two simulator exposures the
drivers completed two surveys the Presence Question-
naire (Witmer et al. 2005) and the IGroup Presence
Questionnaire (Schubert et al. 2001). The surveys were
selected to measure the concept of presence in two
ways, through introspective self-report and causal
factor self-report measures to gain data on the
e!ectiveness of both the wide-screen and the cab-based
simulators in generating immersion. The IPQ measures
spatial presence by asking questions that require
participants report on their own individual experience.
The PQ explores four factors thought to underlie
presence: involvement, sensory fidelity, adaptation and
interface quality.
5. Results
Results were converted to an absolute percentage error
score allowing scores across simulator types to be
compared as seen in Table 1.

35. 5.1. Repeated practice
The comparison between the 2006 and 2007 Trial 1
results indicated that total driver error rates were
higher in 2007 (11.7%) than 2006 (5.2%). However, in
2007, the opportunity for repeated practice resulted in
improved error rates. A 2 (simulator) 6 4 (decision
point) 6 2 (trial) mixed ANOVA was performed on
the 2007 survey data and revealed two significant main
e!ects, one for trial (F ! 13.84; df ! 1, 31, p ! 0.001)
and the other for decision points (F ! 17.29; df !
1.31, p ! 0.000). The main e!ect of trial revealed that
the error rate was significantly greater during the
driver’s first exposure (Trial 1) than during their
second exposure (Trial 2) (11.7% vs. 2.6%, respec-
tively). Therefore, when given the opportunity for a
second exposure in 2007 driver’s error rates decreased
significantly.
5.2. Error rates by simulator type in 2006
The total error rates by simulator were then compared
to determine whether the choice of simulator altered
the e!ectiveness of the training.
In 2006, 12 participants were trained on the wide-
screen while 51 were trained on the smaller cab-based
simulator. A 2 (simulator) 6 4 (decision points) mixed
ANOVA was performed and revealed to be significant
(F ! 55.39, df ! 1. 61, p ! 0.018) for the main e!ect
of simulator. This result indicates that participants had
greater error rates when using the 160 degree curved
wide-screen (9%) than the flat cab-based screen (4%)
over the four major decision points investigated. This
result was interesting in that the e!ect of simulator

36. held over all sections of the virtual track. At each point
in the scene when decisions had to be made to avert
accidents or other incidences drivers using the wide-
screen consistently made more errors than their
colleagues using the cab-based screen as seen in
Table 2. The e!ect of simulator was not influenced
by the type of problem being negotiated on the virtual
track. No significant e!ect was found for the main
e!ect of decision points.
5.3. Individual error types
In terms of specific errors, in 2006, the highest error
rates were all recorded in the wide-screen simulator.
The highest number of errors was recorded (58%)
when drivers were required to report the Children on
Table 1. Error rates for curved and flat screen simulators.
2006 2007 (Trial 1) 2007 (Trial 2)
CS (n ! 12) FS (n ! 51) CS (n ! 4) FS (n ! 30) CS (n ! 4) FS (n !
30)
Average error rate 3.4 1.5 4.75 4.2 2.25 1.4
Error rate as % of total number of actions 9.5 4.2 13.2 11.6 6.25
2.0
Total % (simulator totals combined) 5.2 11.7 2.6
CS, curved screen; FS, flat screen.
462 J.G. Tichon and G.M. Wallis

37. Track incident to the guard after the signaller had been
advised via radio. The second highest error rate (50%)
was recorded again in the wide-screen simulator during
the absolute signal at stop incident, during which the
driver is required to either advise the signaller to
standby or wait to answer the radio after the train is
stopped. The third highest error rate, again recorded
using the wide-screen, occurred in the Failed Level
Crossing incident (42%).
5.4. Error rates by simulator type in 2007
Because of a simulator malfunction a!ecting some
drivers, in 2007 it was not possible to test the
significance of di!erences between the two simulators
in that year. Nonetheless, the general trend was
consistent with the 2006 results, i.e. that the cab-based
screen was producing better training outcomes than
the wide-screen. In addition to making fewer errors on
both Trial 1 and Trial 2, the drivers using the cab-
based screen also made larger improvements between
the two trials as seen in Table 1.
However, the 2 (simulator) 6 4 (decision point) 6
2 (trial) mixed ANOVA performed did reveal a
significant main e!ect for decision points (F ! 17.29,
df ! 1.31, p ! 0.000). The main e!ect of decision
points indicates that almost all of the error rates were
all significantly di!erent from each other which means
that the error rate for the Track Authority Work
condition (24%) was significantly higher than the error
rate for the Children on Track decision point (21%)
which was significantly higher from the Failed Level
Crossing decision point (10%) but Failed Level
Crossing decision point error rate was not significantly

38. di!erent from the error rate in the Absolute Signal at
Stop decision point (8%).
5.5. Presence
Despite the di!erence in skill levels achieved using the
cab-based screen versus the wide-screen; the drivers
rated the two simulators as equal on the presence/
immersion measures. Both simulators scored 69% on
the PQ and 61% on the IPQ as seen in Table 3.
Additionally, both simulators were rated equally by
drivers across all of the four factors underlying the PQ.
There was a greater range between the highest and
lowest scores for both questionnaires received for the
cab-based simulator. However, the greatest deviance of
scores for individual factors occurred with involvement
and adaptation for the cab-based which was also the
case for the wide-screen.
6. Discussion
In terms of the overall training program, the findings
revealed performance did improve significantly when
drivers were given the opportunity to repeat their
simulator experience on the same day. This suggests
that providing a repeat training session in the
simulator may have had a cumulative e!ect on training
e!ectiveness. Practice should result in greater con-
fidence and improved performance. However, the
results also indicate that, although performance
improved across time within a day, retention over
longer periods was not so good. Despite a year’s
additional driving experience in the interim, drivers
returning in 2007 performed significantly worse than
during their training in 2006. Clearly some of the skills
tested in the simulator involve circumstances which the

39. drivers would rarely if ever experience in the real
world, and this lack of repeated exposure may explain
Table 2. 2006 decision point error rates by simulator.
Decision points
Curved screen
(n ! 12) (%)
Flat screen
(n ! 51) (%)
Track work authority 9 7
Failed level crossing 8 4
Children on track 11 5
Absolute signal at stop 10 3
Total *9 *4
*significant di!erence at p 5 0.05.
Table 3. Presence scores for curved and flat screen simulators.
Simulator
Flat screen
Mean SD
Curved screen
Mean SD
Highest possible
score
Flat screen
(%)

40. Curved screen
(%)
PQ (Total) 121 18.5 121 8.5 175 69.1 69.1
PQ factors:
Involve 51.7 9.28 51.7 4.04 77 67.1 67.1
Sensory 18.3 4.69 18.3 1.53 28 65.35 65.35
Adaptation 35.7 5.77 36.3 5.03 49 72.8 74
Interface 15 2.21 14.3 1.53 21 71.4 68
IPQ 51.4 9.02 52 4.58 84 61.2 61.9
Behaviour & Information Technology 463
why some of the skills degraded over time but does not
explain why they degraded to a point below the 2006
level.
Drivers were given a standard 5 min warm-up
before each simulator exposure so it is unlikely that the
di!erence in skill levels attained were a result of
di!ering opportunities to access shorter or longer
warm-up times. However, there may be a number of
explanations for this apparently counter-intuitive
result. It may be that classroom revision time under-
taken prior to the simulation test di!ered dramatically
from one year to the next. This time was not examined
as part of this study. If revision was more targeted one
year to the exact problems the drivers would encounter
during their simulation session this would have
provided much better preparation time. Alternatively,
perhaps the fact that drivers were able to self-select the
simulator they felt most comfortable with and the
driving scene they undertook was unchanged from the

41. previous year, this resulted in over-confidence which
translated into risk-taking in the simulator and
consequently more errors and mistakes. In their second
attempt in 2007, the drivers demonstrated they did
have high skill levels. This begs the question as to
whether the improvement was a product of the practice
in the first trial or whether in fact the drivers were more
focused as a result of having made far too many errors
on their first attempt.
The role of spaced repetitions in learning theory
may also provide a future research direction to explore
in an attempt to explain why skills decayed. There is
very little research on how procedural learning takes
place in simulated environments. However, there has
been a great deal of research on how di!erent spacing
of repetitions in time a!ects the strength of memory
and how the resulting findings could be applied in the
practice of e!ective learning (Wozniak 1995). Memory
can be strengthened through appropriate spacing time
intervals; however, programs must be careful the
spacing of training repetitions does not leave so much
time between sessions that the learned memory trace
becomes completely inaccessible. This theory would
suggest it is highly probable that altering the training
program in a simulator may assist to address the skill
decay problem. In future research, it would be
interesting to investigate how repetition after varying
time intervals either improve or degrade retention
specifically when simulators are in use. In large
companies when relocating sta! to central simulators
is expensive it is particularly important to determine the
longest inter-repetition interval that avoids retrieval
failures.
When comparing learning e!ects across simulator

42. type, the cab-based simulator produced higher decision
making performance scores than the larger wide-screen
simulator. This was surprising as it is generally
accepted that higher-end immersive interfaces are
more successful in rendering scenes similar to real-
world settings. It has been high-end immersive facilities
that have demonstrated that simulation works. How-
ever, researchers who have been working to increase
understanding of immersion and identify the compo-
nents that produce them acknowledge there is still far
to go (Bowman and McMahan 2007). So what might
the reasons be for this apparently counterintuitive
discrepancy? Is it simply the case that occasionally less
is more? Possibly, a smaller screen and smaller
simulator might help aid concentration and reduce
distraction. On the other hand, the cab-based simula-
tors are in one sense closer to a real train. Their narrow
field of view provides a visible frame of reference,
something lacking in the large curved screen simulator.
It may be that this frame of reference contributes to the
realism and fidelity of the simulation, albeit covertly
(as these di!erences did not emerge in the presence
questionnaires).
Equally so it could also be that the perceptual field
scanning for the wide-screen simulator is larger than is
ever done in a real train, and drivers have neither the
time, under the pressures of training for degraded
conditions, nor the experience to scan e!ectively a
‘larger than life’ presentation. The rail training provider
had installed wide-screen simulators to enable teams of
people to be trained in team communication scenarios.
However, the aim of multi-use for the simulators may
have resulted in a detrimental impact for the solo driver
training. A high perceptual immersion achieved via the

43. full visual field may have been achieved at the cost of
validity where the actual visual experience does not
represent the real-world setting.
Both types of simulators used in the study achieved
relatively low presence scores indicating drivers did not
experience the virtual world as highly engaging. One of
the main issues that drivers had was that they ‘. . .had
no sense of train momentum and no feel of going up or
down grades’. Drivers reported being able to adapt
quickly to the controls and interface but the discre-
pancy between how natural the simulator felt in
comparison to a real train may have been significant
enough to reduce the immersive impact of the
experience for some drivers.
Beyond specific issues of simulator design, the
results of this study have highlighted a potential
problem with the use of presence questionnaires in
predicting training outcomes. Despite the significant
di!erence in number of errors made when training was
undertaken in the cab-based simulator versus the
wide-screen simulators, the drivers rated the two
simulators as equal on the presence/immersion mea-
sures. The presence analysis indicates that in terms of
464 J.G. Tichon and G.M. Wallis
subjective feedback neither the cab-based nor the wide-
screen simulators were experienced as significantly
more or less immersive or engaging. The main
di!erence between them was that the scores for the
cab-based simulator were more variable and this
system was rated slightly more highly in terms of the

44. interface which it o!ers.
In terms of having a stress inoculation e!ect, in the
absence of high levels of presence, it is unclear if skill
improvements between training sessions on the same
day were a product of increased confidence in coping
with high stress events or whether it was merely the
result of an increasing level of comfort in driving
through that particular scenario. Retesting drivers
using a new track route along which the same
problematic events (decision points) are embedded
but in a di!erent sequence would reveal whether
drivers do maintain skill levels when under stress but
without the benefit of familiarity.
7. Future directions
In the future it would be interesting to see to what
extent the visual reference frame is responsible for the
discrepancy in performance between the two simula-
tors. It would also be helpful to test alternative
measures of immersion and presence to see if they
correlate more accurately with the training outcomes
described here. The low presence levels did not
predict the good training outcomes achieved on the
simulators. In addition to enhancing performance,
presence is also purported to be a strong indicator of
whether training will ultimately transfer to the real
world. If the higher performance scores are being
influenced by levels of familiarity with the scene it
could be that the lower presence scores should be
more closely studied as an indicator of training
transference. To investigate the e!ectiveness of the
simulator training in the real world one approach
would be to compare the annual incident rates for the
year before training and the year after training for

45. those drivers who have undergone simulation train-
ing. Ideally, a control group of drivers who did not
receive any simulator training would also be com-
pared against those who have. Similarly, incident
rates of drivers who trained on the flat cab-based
versus the curved wide-screen simulators could be
investigated to determine if simulator type has an
impact in the real world.
There is an important role for interactive simula-
tors in replicating degraded events and establishing
them as a core component of rail training programs.
However, any simulator of any level of complexity can
only be as good as the educational program in which it
is embedded. Many simulators are underutilised due
to failure to integrate them into a curriculum that
enhances their potential. Future research investigating
the optimal repetition spacing for simulator training
sessions would contribute greatly to knowledge on how
long memory traces built from virtual experiences last
before requiring refresher training to ensure they
remain accessible to personnel relying on the training
to maintain safe operations.
Acknowledgements
The financial support of the Australian Research Council is
gratefully acknowledged.
References
Baker, D.P., et al., 2005. Medical teamwork and patient
safety: the evidence-based relation. AHRQ Publication
No. 05-0053. Rockville, MD: Agency for Healthcare
Research and Quality.

46. Bowman, D.A. and McMahan, R.P., 2007. Virtual reality:
how much immersion is enough? Computer, 40 (7),
36–43.
Cannon-Bowers, J.A. and Salas, E., 1998. Individual and
team decision making under stress: theoretical under-
pinnings. In: J.A. Cannon-Bowers and E. Salas, eds.
Making decisions under stress: implications for individual
and team training. Washington, DC: American Psycho-
logical Association, 17–38.
Driskell, J.E. and Johnston, J.H., 1998. Stress exposure
training. In: J.A. Cannon-Bowers and E. Salas, eds.
Making decisions under stress: implications for individual
and team training. Washington, DC: American Psycho-
logical Association, 191–217.
Ho!man, H., et al., 2001. Virtual reality monitoring:
phenomenal characteristics of real, virtual and false
memories. Cyberpsychology and Behavior, 4, 565–572.
Hulse, L.M. and Memon, A., 2006. Fatal impact? The e!ects
of emotional arousal and weapon presence on police
o"cer’s memories for a simulated crime. Legal and
Criminological Psychology, 11, 313–325.
Maschuw, K., et al., 2008. The impact of self-belief on
laparoscopic performance of novices and experienced
surgeons. World Journal of Surgery, 32, 1911–1916.
Nanyue, A.Y., et al., 2006. Modeling of virtual tra"c
environment for driving simulator. In: Proceedings of
7th international conference of computer-aided industrial
design and conceptual design, 7–9 November 2006.
Hangzhou, China: IEEE Press, 1–5.

47. Romano, D.M. and Brna, P., 2001. Presence and reflection
in training: support for learning to improve quality
decision-making skills under time limitations. Cyberpsy-
chology and Behaviour, 4, 265–277.
Salas, E., et al., 1999. Team training in virtual environ-
ments. In: K.M. Stanney, ed. Handbook of virtual
environment. Mahwah, NJ: Lawrence Erlbaum Associ-
ates, 873–892.
Schubert, T., Friedmann, F., and Regenbrecht, H., 2001. The
experience of presence: factor analytic insights. Presence:
Teleoperators and Virtual Environments, 10, 266–282.
Stetz, M., Wiederhold, B.K., and Wildzunas, R., 2006.
Usefulness of virtual reality stress inoculation training
for military medical females. Cyberpsychology and
Behavior, 9, 719–720.
Behaviour & Information Technology 465
Thomas, M.J.W., 2003. Operational fidelity in simulation-
based training: the use of data from threat and error
management analysis in instructional systems design. In:
Proceedings of SimTecT2003, 26–29 May 2003. Adelaide,
Australia: SIAA Press, 91–95.
Tichon, J., et al., 2003. Education and training in
virtual environments for disaster management. In:
Proceedings of the World Conference on Educational
Multimedia, Hypermedia, Honolulu, Hawaii, 1191–
1194.
Wallis, G., Tichon, J., and Mildred, T., 2007. Speed

48. perception as an objective measure of presence in virtual
environments. In: Proceedings of SimTect2007, Brisbane,
Australia, 527–531.
Walshe, D., et al., 2005. Virtually driving: are the driving
environments ‘real enough’ for exposure therapy with
accident victims? An exploratory study. Cyberspychology
and Behavior, 8 (6), 532–537.
Witmer, B.G., Jerome, C.J., and Singer, J., 2005. The factor
structure of the presence questionnaire. Presence: Tele-
operators and Virtual Environments, 14, 298–312.
Witmer, B.G. and Singer, M.J., 1998. Measuring presence in
virtual environments: a presence questionnaire. Presence:
Teleoperators and Virtual Environments, 7, 225–240.
Wozniak, P.A., 1995. Economics of learning. Doctoral
Dissertation. University of Economics, Wroclaw.
Zakay, D. and Wooler, S., 1984. Time pressure, training and
decision e!ectiveness. Ergonomics, 27, 273–284.
466 J.G. Tichon and G.M. Wallis
Copyright of Behaviour & Information Technology is the
property of Taylor & Francis Ltd and its content may
not be copied or emailed to multiple sites or posted to a listserv
without the copyright holder's express written
permission. However, users may print, download, or email
articles for individual use.

49. Nick Yee
Department of Communication
Stanford University
[email protected]
Presence, Vol. 15, No. 3, June 2006, 309 –329
© 2006 by the Massachusetts Institute of Technology
The Demographics, Motivations,
and Derived Experiences of Users
of Massively Multi-User Online
Graphical Environments
Abstract
Online survey data were collected from 30,000 users of
Massively Multi-User On-
line Role-Playing Games (MMORPGs) over a three year period
to explore users’
demographics, motivations, and derived experiences. Not only
do MMORPGs ap-
peal to a broad age range (Mage ! 26.57, range ! 11– 68), but
the appeal is strong
(on average 22 hours of usage per week) across users of all ages
(r ! –.04). An
exploratory factor analysis revealed a five factor model of user
motivations—
Achievement, Relationship, Immersion, Escapism, and
Manipulation—illustrating the
multifaceted appeal of these online environments. Male players
were significantly
more likely to be driven by the Achievement and Manipulation
factors, while female
players were significantly more likely to be driven by the
Relationship factor. Also,

50. the data indicated that users derived meaningful relationships
and salient emotional
experiences, as well as real-life leadership skills from these
virtual environments.
MMORPGs are not simply a pastime for teenagers, but a
valuable research venue
and platform where millions of users interact and collaborate
using real-time 3D
avatars on a daily basis.
1 Introduction
Selling virtual weaponry and real estate for a living,
coordinating fifty
people in a dragon-slaying expedition over a period of 5 hours,
marrying
someone you’ll never meet, and switching gender for several
hours at a time.
These are a few of the myriad of virtual phenomena that occur
every day in
online digital constructs known as MMORPGs—Massively
Multiplayer Online
Role-Playing Games. Every day, millions of users (Woodcock,
2004) partici-
pate in these online environments. The increasing prevalence of
these environ-
ments makes it crucial to understand the ways in which we use,
interact, and
live in these digital constructs.
Although many of the theoretical implications of social
interaction in collab-
orative virtual environments (CVE) have been explored in the
artificial con-
fines of virtual reality (VR) research laboratories (Bailenson &
Yee, 2005;

51. Zhang & Furnas, 2002; Bailenson, Beall, & Blascovich, 2002;
Slater, Sadagic,
Usoh, & Schroeder, 2000; Normand et al., 1999; Leigh,
DeFanti, Johnson,
Brown, & Sandin, 1997; Mania & Chalmers, 1998), MMORPGs
provide a
naturalistic setting where millions of users voluntarily immerse
themselves in a
graphical virtual environment and interact with each other
through avatars (vi-
Yee 309
sual representations of users in a digital environment)
on a daily basis. The opportunity to study what people
actually do when they choose to be in a virtual environ-
ment with thousands of other people cannot be over-
stated, and that is the underlying rationale for the cur-
rent studies.
Existing research on computer or video gaming tends
to focus on two main areas: the negative effects of play-
ing video games and whether video games can be used
for pedagogical purposes. For example, an extensive line
of research has focused on demonstrating that violent
video games increase real-life aggression (Ballard &
Lineberger, 1999; Anderson & Dill, 2000; Anderson &
Bushman, 2001), and survey studies have shown posi-
tive correlations between preference for video games
and aggressiveness (Scott, 1995; Funk et al., 2002),
delinquency (Anderson & Dill, 2000), and lower per-
ceived self-concept (Funk, Buchman & Germann,
2000); however, the generalizability of short-term mea-
sures of aggression in experimental studies and the in-

52. ability to infer causality in survey studies is a concern in
this line of research (Griffiths, 1999). One notable ex-
ception is a longitudinal experimental study conducted
by Williams and Skoric (2005). The findings of this
study did not support the assertion that playing a vio-
lent video game will cause substantial increases in real-
world aggression.
Another line of research has explored the pedagogical
uses of video games among elementary school students
(Ko, 2002), high-school students (Ravenscroft &
Matheson, 2002), and college students (Moreno &
Mayer, 2002). Video games can also enhance sensori-
motor tasks (Fery & Ponserre, 2001), visual acuity and
attention (Green & Bavelier, 2003), as well as aid in the
recovery of motor skills after physical trauma (Taylor &
Berry, 1998). In fact, some have explicitly argued that
video games should be considered as platforms for
learning (Gee, 2003).
1.1 What are MMORPGs?
Both lines of above research have mostly relied on
single-player games, but as the ubiquity of high-speed
internet connections and powerful computer processors
have increased with a steady decrease in cost, the para-
digm of computer gaming has changed dramatically,
and MMORPGs are the vanguard of a new generation
of computer games that takes advantage of the accessi-
bility of the internet and the graphical processing capa-
bility of standard computer systems.
MMORPGs are a new paradigm in computer gaming
(see Table 1). By definition, MMORPG users are part
of a persistent world of up to 2000 other concurrent
users (Sony Online, 2003). A persistent world is a world

53. Table 1. Attributes of Three Gaming Paradigms
Attribute Stand-alone games
Local and wide area
network games MMORPGs
Exemplars Solitaire, Snood,
SimCity, Risk
Diablo II, Unreal, Age of
Empires
EverQuest, Star Wars
Galaxies
Cost for player Software Software Software ! subscription
No. of players in world 1 1–16 0–2000
Persistent world No No Yes
Mode of user agency Direct/godlike control One or several
avatars Personal avatar
Size/scope of world Abstracted game board Abstracted or
limited worlds Naturalistic worlds/galaxies,
not abstracted
Player social interaction None Combat strategy driven Rich,
collaborative, social
interactions
310 PRESENCE: VOLUME 15, NUMBER 3
that exists independent of the users. In stand-alone
games and local network games, the world only exists

54. when the game is started by the user, and thus is depen-
dent on the user “activating” it. In an MMORPG, the
world exists before the user logs on, and continues to
exist when the user logs off. More importantly, events
and interactions occur in the world (driven by other
users) even when the user is not logged on to the persis-
tent world. To accommodate the sheer number of users,
the worlds in MMORPGs are vast and varied (in terms
of terrain, flora, fauna, and local inhabitants). In con-
trast, the worlds of most stand-alone and local network
games are simplistic and can only accommodate fewer
than 16 concurrent players in a space that can be tra-
versed in a few minutes.
On a simplistic level, MMORPGs could be thought
of as a scenic chat room with a variety of interactive
tasks. Users experience cities, jungles, and even the fall-
ing rain or snow in rich real-time 3D graphics, and
communicate with each other using typed chat and
templated gestures and expressions. They interact with
the world through a combination of mouse-driven in-
terfaces and typed commands, and partake of a large
number of varied activities that increase in complexity,
reward, and time involvement which typically operate
on a random-ratio reinforcement schedule. These activi-
ties revolve around character advancement and translate
into a functional advantage in terms of the mechanics of
the world, whether this is combat capability, social sta-
tus, avatar appearance, geographic knowledge, equip-
ment quality, or even cooking skills. Whereas the first
few MMORPGs focused heavily on combat-oriented
advancement, recent MMORPGs have offered more
diverse forms of advancement. For example, in Star
Wars Galaxies, one can become a skilled musician, chef,
hair stylist, animal tamer, or politician.

55. Most forms of advancement in MMORPGs require
increasing cooperation or dependency on other users,
oftentimes mutually beneficial. In Star Wars Galaxies,
scouts hunt and harvest hides and meat from animals
which they can sell to artisans who need those resources
to craft basic items. Most combat professions need the
complementary support of each other as they tackle
more and more difficult creatures or enemies that in
turn hold larger rewards. But ultimately, each user de-
cides which form of advancement they will pursue, and
the richness and complexity of the environment elimi-
nates the need for super-ordinate goals or storylines.
Every user is motivated by a different combination of
the possible rewards. The result is that adventures, sto-
ries, and most importantly for the purpose of the cur-
rent work, meaningful relationships between users
emerge during interaction. Functional constructs within
the environment facilitate these social networks— com-
bat groups (temporary collaboration between a few us-
ers), guilds (persistent user-created membership organi-
zations), and ideological alliances (agreements between
guilds or “racial” groups).
Given how different MMORPGs are from stand-
alone and local network games, perhaps a better com-
parison is with live-action or table-top role-playing
games (RPGs) or multi-user domains (MUDs)—the
textual predecessors of MMORPGs. MMORPGs are
different from RPGs in one important aspect. While the
world of an MMORPG is persistent and exists indepen-
dent of players, the world of an RPG exists only when
the players have convened in a physical location. Also,
this physical constraint means that it is not possible to
convincingly change one’s representation in an RPG the
way it is possible in an MMORPG— especially in terms

56. of gender and race. Furthermore, the physical con-
straints of RPGs make it more likely that RPG players
know each other when compared with MMORPG play-
ers. MMORPGs are in fact much more similar to
MUDs than other genres of video games in that both
have persistent game worlds where players can interact
using avatars.
Currently, there is very little quantitative research on
MMORPGs or MUDs. While researchers like Turkle
(1995), Bruckman (1993), and Bruckman and Resnick
(1995) have contributed a great deal to our understand-
ing of the users of MUDs, their approaches were more
qualitative and relied on anecdotal stories, interview
data, or personal experience. For example, Turkle illus-
trated how digital self-representation in MUDs allowed
users to understand the fluid, dynamic, and postmodern
Yee 311
nature of their identities. Almost no quantitative studies
of MUDs exist. One of the few quantitative studies of
MUDs explored how sociability and skepticism toward
computer-mediated communication influenced social
behavior within MUDs (Utz, 2000). Also worth noting
is a multi-methods study of LambdaMOO by Schiano
and White (1998) that illustrated, among other things,
that LambdaMOO users preferred stable, integrated
identities as opposed to the fluid, fragmented identities
Turkle suggested. Existing research on MMORPGs has
mostly relied on analysis of publicly available data. For
example, Castronova (2002) has used the eBay sales
transactions of virtual items to show that the economy
of EverQuest is quantifiable and has a higher gross do-

57. mestic product than some developing countries, and to
show the inequity between the value of the virtual male
and female body (Castronova, 2003). Griffiths, Davies,
and Chappell (2003) aggregated online poll data at
websites catering to EverQuest players to provide the
basic demographics and preferences of EverQuest play-
ers. Very few studies have used primary sources of data.
One of the few is a study by Axelsson and Regan (2002)
that explored the impact of group affiliation on social
behavior in the MMORPG Asheron’s Call. The study
found that group affiliation makes people more social
both online and offline.
Over the course of three years, I have collected online
survey data from 30,000 MMORPG players. Prelimi-
nary qualitative data from open-ended questions were
used to form theoretical questions about the motiva-
tions and relationship formation of MMORPG users,
among other issues (Yee, 2006). In that previous paper,
I argued that the architecture of these environments
facilitates relationship formation, and they are windows
into and catalysts in existing relationships in the material
world. The goal of the current work was to complement
these previous findings by providing a more rigorous
quantitative analysis on who uses MMORPGs, what
motivates their use, and the salience and impact of the
experiences that emerge in these environments. More-
over, this paper attempts to articulate the many oppor-
tunities to study social identity, social interaction, and
relationship formation in these environments.
1.2 Demographics and Usage
The apparent focus of existing video game re-
search on adolescent users creates the illusion that video
game players are a youth subculture, and that video

58. games are a teenage pastime of no important conse-
quence apart from their ability to increase real-life ag-
gression. For example, the entire volume of a recent
special issue of the Journal of Adolescence (Vol. 21-1)
was devoted to the negative effects of video games on
adolescent gamers. The following quotes from that issue
illustrate the assumption that adolescents are the pri-
mary consumers of video games or that video games
somehow impact adults in an entirely different way that
isn’t worth mentioning or studying.
Video games have become one of the favorite activi-
ties of American children. (p.5)
The rise and popularity of video and computer
games as a leisure phenomenon has become an ever-
increasing part of many young people’s day-to-day
lives. (p.1)
A lot of youths are playing violent video games for
many hours per week. When large numbers of youths
(including young adults) are exposed to many hours
of media violence (including violent video games),
even a small effect can have extremely large societal
consequences. (p.120)
In spite of the fact that the average age of com-
puter and video game players is 30 (Entertainment Soft-
ware Association, 2005), the articles in the special issue
seem to perpetuate the assumption that mainly children
and adolescents play video games. In fact, studies in
video game violence in general have mainly focused on
adolescent gamers (see Griffiths, 1999 for review). This
stereotype is also described by other researchers. For
example, Griffiths, Davies, and Chappell (2003) note
that “the image of a typical gamer is seen as socially

59. negative and remains firmly within a youth subculture”
(p.81).
In a recent study that challenged this stereotype
(Griffiths, Davies, & Chappell, 2003), the demograph-
ics and usage patterns of MMORPG users were ex-
plored by analyzing online poll data on two websites
312 PRESENCE: VOLUME 15, NUMBER 3
catering to players of the MMORPG EverQuest. Those
data showed that the game clientele was very much an
adult profile rather than the stereotypical adolescent
player. Although the online polls consisted of large sam-
ples (upwards of 10,000 responses), a weakness with the
study is that the data analysis was bound by the range
and number of answer choices used in the third-party
polls. For example, instead of allowing open-ended re-
sponses to age or hours played per week, only several
ranges were offered. Furthermore, because each ques-
tion was asked on a separate poll, it was not possible to
explore demographics and preferences in relation to
each other for each individual user—such as exploring
age and gender differences in usage patterns. Therefore,
it was crucial to perform a more extensive analysis of the
demographics and usage patterns of MMORPG users.
1.3 Motivations for Use
Articulating motivational differences among differ-
ent users is the precursor to understanding the emer-
gence of more complex behaviors and interactions in
these environments, as well as providing a framework to
differentiate one user from another. To ignore these

60. individual motivational differences is tantamount to
claiming that all MMORPG users are motivated by the
exact same reasons. Most video game studies to date,
however, have relied on the traditional effects model
and don’t take into account the fact that people choose
the media they consume and the varied reasons for do-
ing so (Sherry, 2001; Sherry & Lucas, 2003).
Without an empirical framework with which to identify
individual motivational differences among MMORPG
users, it is impossible to meaningfully differentiate users
or understand their interactions with other users in the
world. This framework provides the foundation to ex-
plore whether different sections of the demographic are
motivated differently, and whether certain motivations
are more highly correlated with usage patterns or in-
game preferences or behaviors.
There have been no systematic attempts to create a
motivational framework for MMORPG users, but an
exploratory framework for MUD users has been pro-
posed by Bartle (1996). Bartle’s proposed “player
types” are derived from his experience in creating and
managing these online textual worlds rather than empir-
ical data, and they provide valuable insight as well as a
framework to test and build upon. Bartle proposed four
types—Achievers, Socializers, Explorers, and Killers—
each having different in-game preferences and motiva-
tions for using the MUD environment. For example,
Explorers are users who are interested in understanding
the mechanics and rules of the system as well as map-
ping out the world, while Socializers are users who en-
joy chatting, interacting, and role-playing with other
users.

61. Even though it is important to be able to differentiate
the motivations among MMORPG users, there has
been no empirical attempt to identify what those moti-
vations might be. The present study attempted to create
an empirical framework for understanding individual
motivational differences among MMORPG users using
an exploratory factor analysis. Also of interest was how
these motivational differences varied across different
demographic sections and how they correlated with us-
age patterns.
1.4 Derived Experiences
The impact that MMORPGs have on their us-
ers—in terms of social interactions, emotional invest-
ment, and acquisition of social skills—will be collectively
referred to as derived experiences in this paper. Even
though there exists very little empirical research on
MMORPGs, there are many reasons to expect that
complex social interactions and social phenomena
emerge in these environments. Indeed, the literature in
MUDs has abundant examples of how intimate relation-
ships and emotionally salient experiences derive from
even textual online environments. For example, Turkle
(1995) has documented romantic relationships, sup-
portive friendships, and even wedding ceremonies in
MUDs. The incidence of a “cyber rape” in a MUD has
also been documented and widely discussed (Dibbell,
1993). The debate it sparked illustrated the amount of
emotional investment users placed in these worlds.
MMORPGs are MUDs on a massive scale with incredi-
ble visual and behavioral richness. Therefore, MMOR-
Yee 313

62. PGs should foster complex social phenomena and inter-
actions among users.
The literature also suggests several reasons for why
this might occur. For example, Walther (1996) sug-
gested that one of the reasons why hyperpersonal inter-
actions—interactions that are more intimate, more in-
tense, and more salient because of the communication
channel— occur in computer-mediated communication
(CMC) is because participants can reallocate cognitive
resources typically used to maintain socially acceptable
nonverbal gestures in face-to-face interactions and focus
on the structure and content of the message itself. The
message itself then comes across as more personal and
articulate. Indeed, in virtual worlds where we do not
have to constantly worry about how we look and be-
have, we would be able to dedicate more cognitive re-
sources to the message itself. Walther also suggested
that as interactants respond to personal messages with
equally personal and intimate messages, the interactions
intensify through reciprocity. In other words, the pro-
cess provides a positive feedback cycle.
Walther’s themes resonate with McKenna and
Bargh’s more recent work (2000) suggesting four fac-
tors that enable positive social interactions online. First,
people have greater anonymity online. Second, the im-
portance of physical appearance is greatly reduced.
Third, the internet transcends the problems of physical
space and wide dispersion. And finally, users have
greater control over the time and pace of their interac-
tions. Again, all of these factors, except for perhaps the
last one, are present in MMORPGs, and suggest why
enhanced social interactions occur in these online envi-
ronments.

63. Behavioral confirmation may also be at work. People
become what we expect them to be (Snyder, Tanke, &
Berscheid, 1977). Given the literal reality of “knights in
shining armor” and the fact that users can choose to be
as attractive as the world allows, users may become
more friendly and more sincere with each other because
of the heroic attributes their avatars project.
The present study used quantitative survey data to
explore three issues that relate to the salience and im-
pact of experiences and social interactions in MMOR-
PGs. First of all, the significance and salience of the rela-
tionships that form in MMORPGs were examined.
Second, the degree of emotional investment in the envi-
ronment was considered. Finally, whether real-life lead-
ership skills could be acquired in the online environ-
ment was explored through self-report measures.
A series of online surveys was used to study the de-
mographics, motivations, and derived experiences
among MMORPG users over a three year period be-
tween the years 2000 and 2003. During this period,
over 30,000 MMORPG users were surveyed, with ap-
proximately 2000 – 4000 respondents in each survey
phase. In the following sections, the methods used and
data collected on three main aspects of MMORPG use
are presented. First, the demographic composition of
current MMORPG users and their usage patterns are
presented to illustrate the wide appeal of these immer-
sive environments. Second, an exploratory factors analy-
sis of the different motivators of usage is presented. Fi-
nally, the salience of the relationships and emotional
experiences users derive from these environments is ex-
amined.

64. 2 General Methodology
Certain methods were common to all three of the
sections discussed in the previous paragraph. These
common methodologies are presented here.
2.1 MMORPG Selection
The approximate numbers of active subscribers to
existing MMORPGs were publicly available (see Wood-
cock, 2005), and it was usually clear which MMORPGs
comprised the bulk of all MMORPG users. From 2000
to 2003, the following MMORPGs together comprised
approximately 75% of the North America MMORPG
market with regard to share of active subscribers: Ul-
tima Online, EverQuest, Dark Age of Camelot, and Star
Wars Galaxies, as calculated using available subscription
data (Woodcock, 2005; CorpNews, 2004). Therefore,
users of these environments were targeted for this study.
314 PRESENCE: VOLUME 15, NUMBER 3
2.2 Participant Recruitment
MMORPG users were recruited through online
websites known to cater to them. These included the
Lore, Stratics, and IGN Vault Networks that have sub-
sites for different MMORPGs. These websites typically
had a news list on their front page with more recent
listings at the top, and each website provided an option
to submit news and information to the editors of the
news list. A concise overview of the goals and scope of
the study together with a link to the online survey itself

65. would be sent to the editors via the news submission
option provided. Publication of the overview was
never guaranteed, and once the news item was listed,
its duration on the news list was dependent on how
many news items the news list could hold and how
many new news items were added over the course of
the next few days.
2.3 Procedure
MMORPG users who followed the link would be
presented with a brief summary of the intentions and
scope of the project, and would have to follow another
link on that page to the survey itself. After providing
informed consent, respondents answered online surveys
consisting mostly of multiple-choice questions in a bi-
nary or Likert-style scale using radio buttons. Certain
continuous variables such as age or hours played per
week were recorded using text fields. Answers to open-
ended questions were recorded using text area boxes.
While open-ended questions were used in the study,
only data collected from close-ended questions, using
Likert-type rating response options, are presented in this
paper, with the exception of age and hours played per
week, which were collected using open-ended text
fields.
Each survey took about 5–10 minutes to complete.
Respondents were not paid or compensated for their
participation. Because respondents were given the actual
goals of the study and not a cover story and there were
no experimental manipulations, there was no need for
debriefing. Duplicate responses were found and deleted
by comparing IP addresses. The rate of duplicate re-
sponses was trivial— 8 out of 3100 in a recent survey

66. phase.
A new survey was usually publicized every 2–3
months in the years 2000 –2003. Surveys targeted all
MMORPGs of interest simultaneously. The typical re-
sponse rate was 2000 – 4000 for each survey. In each
survey, respondents were asked to provide their email if
they were interested in participating in future surveys.
At the beginning of each survey phase, in addition to
the recruitment at websites, respondents already in the
database were contacted via email to notify them of the
new survey in which they could participate. Methods
specific to each section will be provided in the corre-
sponding section.
3 The Demographics and Usage Patterns
of MMORPG Users
3.1 Method
Survey items implemented through standardized
HTML form items (radio buttons and text fields) were
used to gather responses to basic demographic informa-
tion: gender, age, marital status, occupational status,
hours of usage per week, and whether the user partici-
pated with a family member or romantic partner. Age
and hours of usage per week were the only two survey
items implemented with open-ended text fields, while
all other survey items were implemented with set re-
sponse choices. In the tables and graphs throughout the
paper, the age variable is collapsed into five ranges for
ease of presentation.
3.2 Results and Discussion
The data collected over the three year period of

67. the study will be presented in thematic rather than chro-
nological order. While survey data were collected from
users who were active in different MMORPGs, the fo-
cus of this paper is on the demographics and motiva-
tions of a representative MMORPG user group rather
than on comparing the differences among users of dif-
Yee 315
ferent MMORPGs. As such, between-game differences
will not be presented.1 A different number of respon-
dents participated in every survey phase (typically 2000 –
4000 respondents participated in each survey), and find-
ings are drawn from different survey phases. Therefore,
the sample size for the results, graphs, and tables below
are not all identical. The majority of respondents were
male (85.4%, n ! 5547). The average age of the re-
spondents was 26.57 (n ! 5509, SD ! 9.19); the me-
dian was 25, with a range from 11 to 68. The lower and
upper quartile boundaries were 19 and 32 respectively
(see Figure 1 for the ages of both genders). Thus, only
about 25% of MMORPG users are teenagers and
MMORPGs have cross-generational appeal.
Female players (M ! 31.72, SD ! 10.11, n ! 788)
were significantly older than male players (M ! 25.71,
SD ! 8.73, n ! 4705), t(5491) ! –17.46, p " .001,
r ! .22. While the cause of this age difference may be
hard to pinpoint, one potential explanation lies in how
male players and female players were introduced to
MMORPGs. For female players, 26.9% (n ! 420) were
introduced to the game by their romantic partner (boy/
girlfriend, fiancé/e, or husband/wife), compared with
1.0% of male players (n ! 1778), p " .001. Since peo-

68. ple with romantic partners tend to be older, this form of
recruitment may have disproportionately increased the
average age of female players.
Other findings also demonstrate that MMORPG us-
ers are not primarily adolescent students. It was found
that 50.0% of respondents (n ! 2846) worked full-time,
while another 22.2% were full-time students. A more
detailed breakdown by gender is shown in Table 2. Of
particular interest is that 13% of female players were
homemakers (13.3%, n ! 438), which is a striking jux-
taposition to the stereotypical image of the male teen-
ager; however, it is exactly this coexistence of male teen-
agers and middle-aged homemakers in MMORPGs that
highlight the wide appeal of these environments. Fur-
thermore, a substantial portion of respondents had es-
tablished families of their own. This is demonstrated by
the substantial portion of respondents who were mar-
ried (36.3%, n ! 2846) or had children (22.1%, n !
2846). Thus, the overall demographic composition of
MMORPG users is quite diverse, and does not consist
primarily of adolescents. In fact, it also includes college
students, early adult professionals, middle-aged home-
makers, as well as retirees.
It is important to establish the wide appeal of
MMORPGs for two reasons. The first and foremost is
to counter the stereotype that video gamers are part of a
youth subculture, implying that these online environ-
ments only appeal to a small slice of the general popula-
tion. Second, establishing the diverse demographics of
MMORPG users makes the following data on usage
patterns more poignant in showing the strength of the
appeal of these environments. On average, respondents
spent 22.71 hours (n ! 5471, SD ! 14.98) each week