Content uploaded by Oleg Špakov
Author content
All content in this area was uploaded by Oleg Špakov on Aug 18, 2015
Content may be subject to copyright.
LONG PAPER
Gaze controlled games
Poika Isokoski ÆMarkus Joos ÆOleg Spakov Æ
Benoı
ˆt Martin
ÓSpringer-Verlag 2009
Abstract The quality and availability of eye tracking
equipment has been increasing while costs have been
decreasing. These trends increase the possibility of using
eye trackers for entertainment purposes. Games that can be
controlled solely through movement of the eyes would
be accessible to persons with decreased limb mobility or
control. On the other hand, use of eye tracking can change
the gaming experience for all players, by offering richer
input and enabling attention-aware games. Eye tracking is
not currently widely supported in gaming, and games
specifically developed for use with an eye tracker are rare.
This paper reviews past work on eye tracker gaming and
charts future development possibilities in different sub-
domains within. It argues that based on the user input
requirements and gaming contexts, conventional computer
games can be classified into groups that offer fundamen-
tally different opportunities for eye tracker input. In
addition to the inherent design issues, there are challenges
and varying levels of support for eye tracker use in the
technical implementations of the games.
Keywords Computer games Eye tracking Taxonomy
1 Introduction
Computer games are a significant part of popular culture.
Eye trackers offer interesting human-computer interaction
opportunities for people with and without disabilities. This
paper discusses these opportunities from two points of view
in the context of computer games. Firstly, eye trackers are
widely used to assist people with disabilities that prevent
manual operation of computers. For such users, many of
whom have limited access to collaborative gaming, the
possibility of playing computer games using eye trackers
would be highly desirable. Secondly, eye trackers may
enhance the playing experience for all players, with or
without the usual manual input devices. Mainstream
gaming industries are unlikely to adapt their products sig-
nificantly for the benefit of the minority of users who
currently rely on eye trackers. However, if benefits for all
players were found, mass-market products may be more
likely. People with disabilities would benefit in the form of
increasing availability and decreasing prices of eye track-
ers, while gaming in general would benefit from
incorporating a new and powerful interactive and attention-
aware tool.
At first glance it may seem as though little is needed in
order to make standard computers eye tracking capable.
The most popular operating principle is the video-based
corneal reflection eye tracking [4]. A typical corneal
reflection eye tracker follows the way that light reflects
from the cornea and the retina. To avoid complications due
to light sources outside the tracker, infrared light is usually
used. Not having other intense infrared radiation sources
near the computer is a reasonable assumption in indoors
P. Isokoski (&)O. Spakov
Department of Computer Sciences/TAUCHI,
University of Tampere, 33014 Tampere, Finland
e-mail: poika@cs.uta.fi
O. Spakov
e-mail: oleg@cs.uta.fi
M. Joos
Department of Psychology,
Dresden University of Technology, Dresden, Germany
e-mail: joos@applied-cognition.org
B. Martin
LITA, University of Paul Verlaine, Metz,
57045 Metz Cedex 1, France
e-mail: benoit.martin@univ-metz.fr
123
Univ Access Inf Soc
DOI 10.1007/s10209-009-0146-3
environments where light bulbs are usually in the ceiling.
In addition to the infrared illumination, the trackers consist
of a video camera that is sensitive to infrared light, a CPU
that is powerful enough to analyze the video stream in real-
time, and software that calculates the point where the
user’s gaze intersects the display.
1
All these components,
except the infrared illumination and the eye tracking soft-
ware, are present in many computers sold today. Most
video cameras are sensitive to infrared light. In fact they
are so sensitive that they have a filter which blocks infrared
light. Making this filter adjustable so that it could be made
to block visible light for eye tracker use is probably not
expensive. The cost of adding a few infrared LEDs to the
sides of displays in order to illuminate the user’s eyes is
also rather small. Thus, there appear to be no significant
technical hindrances.
Unfortunately, this is not quite the case. Typical cameras
do not have sufficient resolution or the kind of optics
necessary. Although the addition of viable optics, increased
resolution, and infrared illumination does not necessarily
imply a huge rise in cost, computer manufacturers are
unlikely to add hardware to their products without a sig-
nificant market demand. Therefore, it seems likely that
unless new applications are found, eye trackers will remain
specialized devices and will not become a part of the
standard computer setup. However, if a successful mass-
market product is found, it is likely that unit prices of eye
tracking equipment will decrease, even dramatically, as the
demand for eye trackers brings about larger manufacturing
volumes.
In search for applications that could make large-scale
eye tracker manufacturing a realistic development, this
paper reviews past work in eye controlled games and dis-
cuss possible future developments, concentrating on the
use of an eye tracker as the only input device. It should be
kept in mind that this is an unlikely scenario for the
majority of computer users. Most users today use several
input devices including mice, keyboards and dedicated
game controllers according to the demands of the game and
the context of use. This practice is likely to continue with
eye tracking as a possible addition to the arsenal. Thus, the
point of tight focus on eye tracking is only to highlight the
strengths and weaknesses of eye trackers as gaming devi-
ces, not to propose abandoning other devices or to exclude
multi-modal use of all available input and output devices.
Under a pragmatic point of view, the main interest here are
the technical and practical issues involved in eye tracker
gaming.
Eye tracker gaming is in its infancy. Most researchers
have approached the area from the point of view of existing
games and asked questions like how could eye trackers be
integrated in a particular game. This is the approach
adopted in this paper as well. In the following, two ways of
classifying computer games are presented. Firstly,
according to the type of input that a player must accom-
plish in order to play successfully. This classification
allows identifying games which could be controlled solely
by eye movements without hindering successful gaming.
Secondly, games can be classified according to the means
that must be used to make them work with eye trackers.
This is a pragmatic classification which allows estimating
the skills and effort necessary in order to produce an eye
tracker compatible version of a particular game. Finally,
examples of eye controlled game projects are discussed,
followed by conclusions on the current state and likely
future developments of eye controlled games.
1.1 Eye movements and games
Eyes are primarily sensory organs. While the field of view
of human eyes is wide, both eyes have only a small area
(fovea) that can see accurately. The fovea is circular and
covers about 1–5 degrees of the visual field [4]. The range
from 1 to 5 degrees is not intended to represent variation in
the actual size of the fovea, but rather the varying use of the
term. The accuracy of vision does not change abruptly
from sharp to fuzzy at a certain point. Instead, there is a
gradual degradation when moving further from the center
of the fovea, thus the ambiguity on where to draw the
border.
When perceiving a scene, humans move their eyes in
rapid jerky movements known as saccades to direct the
light from different parts of the scene onto the fovea.
Between saccades the eyes remain relatively fixed. These
stops are known as fixations. In addition to saccades and
fixations, eyes can follow moving objects in a slow smooth
movement known as smooth pursuit. Because receptor
cells in the eyes only react to changes in the amount of
light they receive, and in lack of changes quickly adapt and
stop responding, the eyes need to move during long fixa-
tions. These movements are small and usually not noticed
without special equipment such as eye trackers.
The eyes can be controlled with remarkable speed and
accuracy even in difficult situations, such as under simul-
taneous body and head movements. However, because of
the size of the fovea, the accuracy needed for large vision-
related saccades is not very high. Abrams et al. [1] report
standard deviation of saccade endpoints between 0.7 and
0.9 degrees depending on saccade length. This is under-
standable, since there is a need for rapid eye movements to
be accurate enough to quickly project any given point in
1
Calling such devices gaze trackers might be more accurate, since
their most important function is that they track the eyes with such
precision that they can calculate where the eye is pointing. We call
these devices eye trackers to follow the convention within the field.
Univ Access Inf Soc
123
visible space on the fovea, but accuracy beyond that is
superfluous. Small microsaccades can be used to further
adjust eye position more accurately.
A good eye controlled game might utilize the knowledge
of these naturally occurring eye behaviors—either to by
supporting their use or by challenging the player to learn
new behaviors.
In a game the use of the eyes for observation of the
scene must also be possible at any time. If the natural use
of eyes for information gathering causes undesired actions
in the game, the player may feel distracted and trapped.
Thus, switching between observing and controlling must be
effortless. Preferably, the game should be modeless, so that
observing and controlling happen simultaneously without
conscious effort.
For good results in fast-paced games rapid and fluid use
of eye control is essential. Otherwise, an eye control bound
player will be at a disadvantage against other players with
other input devices. In turn-based games the requirements
for the efficiency of eye control are lower, and inefficient
control methods are less likely to interfere with a satisfying
gaming experience.
These considerations allow to characterizing the degree
to which a game is compatible with eye control. Turn-based
and other games which do not require continuous on-line
control can be easily adapted for use with an eye tracker.
The same is true for many single player games which
require real-time control, because the difficulty of the game
can be adjusted to fit the player’s performance. Games that
require continuous control and are played competitively
against people with other input devices constitute a signif-
icant challenge for eye tracker input design, because other
players and other input devices determine the difficulty
level. Table 1summarizes the situation. It divides gaming
situations into those which require turn-based and contin-
uous control on the one hand, and to solitary and group
situations on the other.
The following sections discuss the relationship of dif-
ferent in-game interaction techniques and the framework
in Table 1. Interaction techniques are intended as the dif-
ferent ways that the players use the input devices to control
the game. Pointing and clicking is perhaps the most com-
mon interaction technique. It is used in different contexts
in different games. For example, in chess it could be the
only interaction technique needed to choose the piece to
move and then the square to move it to. In other games
pointing and clicking may appear as a part of more com-
plex interaction techniques. For example, in a first person
shooter game, the player may be pointing and clicking on a
target to shoot at it, while simultaneously moving to avoid
enemy projectiles. Situations like this where there are
multiple degrees of freedom that must be controlled
simultaneously pose significant challenges to eye tracker
input as the only input technique. In multimodal use of
several input devices these situations offer opportunities
for designs that offer efficient combinations of the input
modalities.
1.2 Methods of implementing eye control in games
On the technical side, there are four different ways of
implementing eye control in games. The simplest of these
requires no game specific modifications. Most eye trackers
offer an eye-mouse mode of operation where the tracker
software commands the operating system to place the
mouse cursor at the point where the user is looking at. Most
trackers also emulate the mouse button press by sending a
button press to the operating system whenever the user’s
gaze remains within a small area for a pre-determined
amount of time. This selection technique is known as
dwell-time based selection. It is a trade-off between acci-
dental selections due to too short dwell-time and slow
operation due to too long dwell-time. The optimal settings
depend both on the context of use and on the user’s pref-
erences. However, together these mouse emulation
techniques are sufficient to play some games.
The second simplest solution is to use additional soft-
ware outside both the eye tracker and the game itself. In a
sense this is what the eye-mouse software bundled with
most trackers does. Sometimes, however, no such generic
solution exists and the external software needs to be tai-
lored for each game separately. For example, a program
which repeatedly sends events corresponding to pressing
the trigger button may be sufficient to make some games
playable with the eye-mouse functionality of an eye
tracker.
Table 1 The challenges in adapting computer games for eye tracker use as a function of control mode and number of players
Control
Turn-based Continuous
One player Easy (chess against a computer) Moderately difficult (FPS against a computer)
Many players Easy(chess against a human) Difficult (FPS against humans)
FPS refers to first person shooter games where the player moves in a three dimensional virtual world constantly controlling the direction and rate
of movement
Univ Access Inf Soc
123
The third possibility is that the source code of the game
is available and can be modified to allow eye control. This
approach tends to be more labor intensive than the two
previous ones. It is also rarely applicable, since the source
code of most commercial games is not available.
The most labor intensive method of all is building a new
game from scratch. Although expensive, this method also
allows the game design to maximize the potential of eye
tracker input. Eye controlled games which represent a
genuinely new gaming experience for all players, and differ
significantly from the dominant game genres, can be
achieved only by creating completely new games.
The level of effort is also a reasonable predictor of the
number of games that are available in each category. There
are a large number of games that can be played with the
basic eye-mouse and text entry functionality available in
most eye tracker packages. There are some projects that
have created add-on software that makes playing of some
games possible
2
and some projects that have modified the
internals of existing games. On the other hand, no com-
mercial projects are known to develop a completely new
game for eye tracker use. Examples of student projects that
have implemented eye tracker controlled game prototypes
can be found on the COGAIN website at http://www.
cogain.org.
2 The state of the art in eye controlled games
Long before the integration of eye trackers in interactive
computer systems, special eye tracking technologies such
as cameras were developed in order to study human per-
ception and cognition through experimentation [2]. Eye
tracking experiments that presented stimuli and recorded
the viewer’s reactions were developed in order to under-
stand various aspects of attention and cognition. Wade and
Tatler [16] provide an impressive overview of the historical
development of eye tracking research and methodology.
Although computerized eye tracking systems were not
involved until much later, the experience of a participant in
these early experiments was sometimes close to a computer
gaming experience. Just like many computer games, these
experiments required fast reactions and sometimes a degree
of problem solving. However, usually there was no story-
line and no intention of immersing the participants in a
game world. These missing aspects are often considered
distinctive characteristics of computer games.
One of the first documented prototypes with
computerized real-time eye tracking and intentionally
constructed storytelling was the gaze responsive self
disclosing display by Starker and Bolt [15]. The display
showed a planet that rotated slowly. On the planet an
observer could see objects such as mountains and stair-
cases. When the observer’s gaze dwelled on these objects
long enough, the system gave more information about the
object of interest using synthesized speech. While such
dynamic story telling systems may not be considered
games by today’s standards, many games have similar
features. Exploring one’s surroundings and interacting
with objects and characters that are found during these
explorations is a frequently used way of revealing the plot
of the game to the player.
Eye trackers have traditionally been expensive and dif-
ficult to use in comparison to other input devices. For
example, traditional eye tracker issues such as restrictions
to head movements and need for calibration are not a part
of the use of most mainstream input devices. Consequently,
the potential market for eye controlled games has been
small. Eye controlled games have so far been research
prototypes or training tools to be used when introducing an
eye tracker to a new user, although therapeutic applications
have also been proposed [11]. In recent years, eye trackers
have developed into a more user-friendly direction and
both the number of games bundled with eye trackers and
the number of research prototype games has increased. In
the following, some of these games are described in order
to give a flavor of the current state of the art.
The Eyegaze communication system by LC Technolo-
gies includes paddle games and a Score Four game in its
software package. The paddle games involve an on-screen
paddle which is used to stop an on-screen ball from hitting
one of the sides of the playing area. This is very similar to
‘PONG’, one of the earliest video games, released 1972.
Score Four is a game where the players alternately place
marks on a grid and the first to have four marks in a straight
line wins. In addition, LC Technologies has developed a
gaze-operated version of Mahjong, an old Chinese board
game.
Eye tracker input is especially suitable for paddle
games. The player’s task is to place the paddle so that it
stops a ball from escaping the playing area. This task
becomes trivial if the paddle follows the player’s gaze. All
the players need to do is to follow the ball with their eyes.
Dorr et al. [3] conducted an experiment to verify that using
eye tracker input instead of a mouse indeed improves
player performance in this kind of a situation. They set up
an experiment with 20 students playing against each other
in pairs. One player played with a mouse defending one
edge of the screen and another with an eye tracker
defending the other edge. Each player in a pair played with
both input devices. The eye tracker had a statistically sig-
nificant advantage demonstrating its suitability to fast-
paced paddle games.
2
For an example see the Chicken shooting game example in this
paper.
Univ Access Inf Soc
123
At the time of writing this article, the MyTobii system
by Tobii Technology included a Minesweeper game and a
Gobblet game. The Minesweeper is an eye controlled
version of the well known strategy game made famous by
its inclusion in the Microsoft Windows operating system.
Gobblet is a four-in-a-row variant played on a 4 94 board.
It deviates from the traditional tic-tac-toe mode of play by
allowing the moving of pieces and having playing pieces of
different sizes. A big piece can be placed on a smaller piece
thus stealing the opponent’s position.
Oleg Spakov has implemented Chess, Tic-tac-toe, and
Lines in the MyTobii environment [13,14]. Chess and tic-
tac-toe are well known games. Lines is a puzzle game
invented by Oleg Demin [17]. The goal is to arrange
similar items into line or cross formations that, when
completed, disappear. The challenge in the game arises
from the fact that new pieces appear on the display after
every move. One must make moves which are likely to
help the formation of multiple shapes to avoid having the
game board getting completely filled with pieces.
Jo
¨nsson [8] experimented with Half Life (First Person
Shooter) and Sacrifice (Shoot-em-up). Her participants
claimed that eye controlled gaming was more fun than
using traditional input devices. Additionally, players
achieved higher scores in Sacrifice with eye control than
without it. Player performance in Half Life was not
reported. Before deciding to use the Sacrifice game for her
experiment, Jo
¨nsson had tried another Shoot-em-up game
where targets moved across the screen. She found that
players tend to track the target which leads to shots landing
slightly behind the target.
The same finding was reported by Smith and Graham
[12], who experimented with three games: Quake 2 (FPS),
Neverwinter Nights (Role playing game), and Lunar
command (Shoot-em-up). Lunar Command is a two-
dimensional shooting game where the player defends bases
at the bottom of the screen from missile attacks by shooting
at the missiles by pointing at the desired interception point
and pressing a button. Smith and Graham reported lower
scores with eye control than with mouse control. They
explained that this is mostly because it is very difficult to
aim at the empty space in front of the target using eye
pointing. Aiming at an intersection of the missiles extrap-
olated trajectory was necessary in Lunar command in order
to compensate for the time of flight of the projectiles.
Quake 2 is a First Person Shooter game where the player
moves in a three dimensional world and shoots at the
creatures found there while avoiding being shot. Never-
winter Nights is a computer-based role playing game where
the player’s avatar is shown from a third person perspective
and moved by pointing and clicking. For both these games
Smith and Graham report lower performance with eye
control than with mouse control. The players received little
training. With training in eye control comparable to their
experience with the mouse the situation might be different.
However, the questionnaire data reported by Smith and
Graham shows that the players felt more immersed in the
game when using eye control. An overall verdict on the
value of eye trackers as gaming devices is difficult to make
based on the evidence gathered by Smith and Graham. This
is because it is difficult to weigh performance against
immersion. If the game does not require competition
against people using other input devices, the value of
immersion may be greater than the value of a high score. If,
however the other players have more efficient input devices
and can effortlessly beat the eye tracker based player,
performance may be felt as more important, no matter how
immersed the losing player was.
A third report on eye trackers in FPS games is from
Isokoski and Martin [6] and Isokoski et al. [5], who
implemented an eye tracker input mode in their input
device testing software. This software looks like a FPS
game to the player, but was written for the purpose of
testing input devices in pointing and navigation tasks in
three dimensional virtual worlds. Isokoski and Martin
were interested in comparing the efficiency of different
input devices. Their comparison included a gamepad
with thumb-operated joysticks, keyboard and mouse, and
a combination of mouse and eye tracker. In the eye
tracker mode, aiming could be done with two crosshairs.
One stayed in the middle of the display and was used for
aiming by rotating the point of view using the mouse.
The other crosshair followed the players gaze. During
the experiment only the eye controlled crosshair was
used in the eye tracker mode. The results suggest that
the eye tracker mode was competitive with the gamepad,
but the mouse and keyboard mode was more efficient
than the two other modes. Isokoski and Martin report
only preliminary results for one player, thus it is difficult
to generalize the results.
Isokoski et al. [5] reported a further study where six
participants played for 50 min in 10 five-minute blocks
using three different input device configurations. This
study used the same environment as the 2006 study testing
three different input device configurations. The configura-
tions were: (1) full gamepad control, (2) moving with a
gamepad and aiming with eyes, and (3) steering and aiming
with eyes (only velocity control and trigger for shooting
were operated manually with the gamepad). The results
showed that the increasing eye control did not affect the
players’ performance in terms of targets hit, but did
increase the number of shots fired.
Generally speaking, all eye tracker gaming experiments
have been short. It is hard to predict what users will think
of eye trackers in the context of computer games in the
long run. Additionally, not much is known about the
Univ Access Inf Soc
123
potential dangers of eye tracker gaming in the form of
repetitive strain injury, for example.
3 Game genres and their challenging features
This section discusses traditional computer games and the
suitability of eye tracker input for controlling them. The
notion of game genre is used in a colloquial way to mean
the often poorly defined and constantly changing loose
grouping of games into categories based on similarity.
Often a genre in this meaning is born when a game is
particularly successful and several similar products appear
as a result. What makes these groups of games poorly
defined is that some contain features that were not present
in the original one. Over time, successful genres tend to
split into several sub-genres that may merge with other
sub-genres creating a mesh that is very difficult to describe
comprehensively. Rather than comprehensive, the current
discussion is intended to be of practical value by high-
lighting properties of games that make them particularly
suitable or unsuitable for the use of eye trackers as input
devices.
Table 2presents a brief overview of the positive and
negative indicators for eye tracker control for each game
genre. The issues identified in Table 1are included (one vs.
many players and real-time vs. turn-based game). In
addition, multiplayer situations present two levels of dif-
ficulties. The first is the need to communicate and perform
well enough to manage in a multiplayer situation (Online
multiplayer), and the second is to do this in a game that is
not turn-based, but requires instant reactions (Online real-
time multiplayer). Games requiring continuous position or
velocity control or activation of controls while visually
attending to something else pose special challenges in
implementing eye tracker input. These, as well as other
aspects of these game genres, are discussed in more detail
in the following sections.
3.1 Computerized board games and puzzles
Board games, as well as their computerized versions, are
usually one player or multiplayer games with turn-based
control. Therefore, they do not place excessive demands on
the efficiency of eye tracker input implementation. For
example, chess, go, monopoly, checkers, and many other
traditional games do not have any inherent hindrances
against successful eye controlled versions. Eye controlled
versions can be implemented by merely making sure that
all user interface elements are big enough for easy selection
with a pointer controlled with an eye tracker. The same
holds for many puzzles. For example, implementing eye
controlled versions of crosswords, Sudoku, and missing
square puzzles does not pose significant challenges for the
user interface design.
Unfortunately, theoretical simplicity does not necessar-
ily mean that all these games are easy to play with eye
trackers in practice. Currently available versions may have
targets which are too small to select comfortably with eye
pointing. Games may also be programmed using platforms
which read pointer events directly from the device, making
it difficult for eye tracking software to feed mouse events
into the game. An example of such problems are games
that are based on the DirectX library and programming
interface by Microsoft; using direct input from keyboard or
mouse through DirectX makes it difficult to inject events
generated by an eye tracker.
3.2 Computerized card games
Card games that are mostly about strategy and do not
require quick reactions are also based on turn-based control
and, therefore, well suited for eye control. Computerized
poker over the Internet, for example, has recently become
very popular. Again, in theory these games should be easy
to operate with eye trackers, but there are practical tech-
nical hindrances. Typical online casinos use software based
on bitmap graphics. Consequently, the user interface is not
resizable. While many in-game controls, such as buttons to
select the next action, are large, the ‘‘lobby’’ interface used
for selecting the game to play and managing the system
settings tends to be full of very small targets making eye
controlled use difficult.
Given suitable client software, eye controlled play
would be possible in such games. Unfortunately, online
casinos typically do not allow third party clients to be used
on their servers. In addition to card games, many other
online casino games are turn-based, with generous time
limits that make them, in principle, suitable for eye con-
trolled play.
3.3 Shoot-em-up (scrolling shooters)
Shoot-em-up games are games with usually two degrees of
freedom position control in addition to a trigger and a
possible weapon/shield switching control. An archetypical
shoot-em-up game has a downward scrolling background
that brings with it targets that appear at the top of the
screen. The task of the player is to maneuver his or her
game character (often a spaceship) around the screen and
shoot as many of the targets as possible while avoiding the
projectiles launched by the targets and other hostile
objects.
Eye control in these games case is challenging, since
shoot-em-up games require constant or at least frequent
control of position. They often also require a dissociation
Univ Access Inf Soc
123
of attention to the target, and position control of the
shooting device. The player needs to look at a target, but
simultaneously stay out of the line of fire if the target
defends itself. Also, when avoiding enemy projectiles the
player needs to steer the avatar through empty areas. The
natural reaction is to look at the approaching projectiles.
Therefore, direct mapping of gaze position to avatar posi-
tion would probably lead to a disaster.
Since shoot-em-up games are usually one player games,
the input efficiency requirements are not necessarily high.
Furthermore, with slight changes to the control require-
ments, eye controlled play can be made quite rewarding.
For example, the need for the trigger can be eliminated by
implementing an auto-fire function. This modification
removes the tactical aspect of conserving ammunition in
the game, but in many games, the maneuvering task alone
can be rewarding.
3.4 Beat-em-up (fighting games)
Typical gameplay in a beat-em-up game consists of well
timed operation of mini joysticks and buttons on a game-
pad. The player’s character in the game uses its limbs and
the weapons it holds to attack its opponents. Such games
usually involve a wide variety of possible attack moves
which must be executed in sequences that penetrate
the defenses of the opponent. The player must also move
the game character around the scene to avoid being hit by
the opponents.
Implementing eye controlled mode is difficult. The large
number of controls to be operated is difficult to map to eye
movements. It may be possible to build a game where one
aims the attacks by looking at the desired part of the
opponent’s body and the correct move sequence for the
attack is then automatically selected. However, there are no
games reported in the literature that implement this kind of
control method. Consequently, it is difficult to tell h a good
gaming experience is achievable. However, it can be
anticipated that this kind of eye control mode changes the
nature of the game considerably. The manual skill involved
in operating the controls is one of the features which attract
players to beat-em-up games. The player learns fighting
skills and enjoys the power that applying those skills gives
in the game world. Automating the fighting may make
the game less attractive. However, it may also enable the
player to be more immersed in the game, and promote the
importance of tactics over manual dexterity.
3.5 First person shooters
First person shooter (FPS) games are a game genre where
the game world is shown from the point of view of the
player’s avatar and where the interaction with the objects
and creatures of the game world happens mainly by
shooting. Traditionally, moving in a three dimensional
world requires at least four degrees of freedom (2DOF for
both translation and rotation) and several discrete controls
such as buttons for shooting, jumping, crouching etc. In
addition, FPS games are often played competitively over
the Internet. Therefore, a competitive eye control imple-
mentation would need to be comparable to skilled two
handed play.
Table 2 Positive and negative indicators for eye tracker compatibility of a game genre
Game genre Indicators for eye tracker use
Positive Negative
One
player
mode
Turn-based
gameplay
Online
multiplayer
Online real-time
multiplayer
Continuous
position control
Dissociation of
focus of attention
and control
Large number
of commands
Board games and puzzles x x x
Card games x x x
Shoot-em-up x x x
Beat-em-up x x x x
First person shooters x x x
Flight simulators x x x x x
3rd person action and adventure x x
Level jumping (platform) x x x
Turn-based Strategy x x x x
Real-time strategy x x x
Turn-based role playing games x x x
Real-time role playing games x x x
Racing x x x
Univ Access Inf Soc
123
A special difficulty is that the ability to aim accurately is
very important in FPS games. If a player can hit far-away
targets accurately, he or she has a significant advantage,
because the distance gives enough time to evade enemy fire
and the likelihood of chaotic close combat situation is
smaller. The preferred aiming device of many FPS players is
the mouse. Eye trackers are not competitive with the sub-
pixel positioning accuracy of modern mice. Aiming is also
problematic when using the thumb-operated joysticks of
game controllers (such as those delivered with Microsoft
Xbox 360, Sony Playstation 2 and others). Some games have
implemented systems which assist with aiming. For exam-
ple, Klochek and MacKenzie [9] describe a game (Halo 2 by
Bungie Studios and Metroid Prime by Retro Studios and
Nintendo) where the game helps the player in maintaining a
good aim even if the target moves. The aiming help needed in
eye-based aiming is different. Acquiring and following tar-
gets with eye-based aiming is easy. Due to the inaccuracy in
fast eye movements, tracker inaccuracies, and tracker noise,
hitting small targets efficiently is difficult. It may be possible
to add some aiming assistance which could be used when eye
tracker FPS performance needs to be competitive with
players using a mouse. In one player situations such systems
may not be necessary.
Navigation in three dimensional spaces shown from the
first person perspective is a special situation of gaze-based
interaction, because players have extensive experience in
real world navigation with the same point of view. Eyes
can be used for controlling the orientation of the game
character in a fairly natural way, by turning the field of
view whenever the player looks away from the center of
the display. When an object of interest becomes visible, the
player naturally tracks it with his or her eyes and the
turning of the field of view could automatically stop when
the object of interest reaches the center of the display.
Overall, however, FPS games cannot currently be played
efficiently using eye tracker input alone. Good solutions for
eye-based velocity control and trigger operation are
needed.
3.6 Flight simulators
There are different kinds of flight simulators. Some are
based on scenarios of realistic navigation of passenger and
cargo planes on rather peaceful routes. The pace of
gameplay is often slow enough to be categorized into the
turn-based category in Table 1. However, simulators that
require continuous control are almost as difficult to use by
eye control as FPS games. There is a whole class of flight
simulators where the difficult flying conditions are an
important part of the game. For example, damaged fighter
planes may need to fly under enemy attacks in a difficult
weather. All flight simulators require at least periodically
observing the environment while manipulating the controls
to make the simulated airplane turn. This is difficult to
achieve with eye tracker input.
In addition, implementing control modes where a lot of
the manual control is automated changes the game. The
whole point of a simulator is to learn and use the skills
needed in the simulated environment. Removing the need
to do so removes something essential from the game. Thus,
it would appear that flight simulators are not a fruitful
avenue to pursue in eye tracker gaming at least if tradi-
tional airplane controls are being simulated and only eye
tracker input is available.
3.7 3rd person action and adventure games
Instead of the first person perspective, many action and
adventure games use a third person perspective. These
games typically relieve the player from the need to control
the camera angle. In first person games the scene is ren-
dered as if it was seen through the game character’s eyes.
Third person games are shown from different angles, often
as if there was a cameraman following the game character.
The camera can be locked into a specific position in rela-
tion to the game character or the game may position the
camera automatically. Depending on the camera angle, the
player can see more of the events around the character than
in games shown from the first person perspective. How-
ever, controlling the game character remains complicated
and it is difficult to map all the controlled degrees of
freedom to eye movements.
Sometimes detailed control over the character’s move-
ments is not necessary. For example, moving can be
completely automated so that the player only needs to point
to where he or she wishes the character to move, and the
game controls the character’s movement to the desired
location, even dealing with the obstacles on the way. The
character navigation in the role playing game example
given by Smith and Graham [12] was of this sort. In the
gaze controlled version the player pointed with eyes and
clicked the mouse button to make the game character move
to the pointed coordinates.
Overall, using eye trackers to control these games is
possible in principle, but most games do not work without
modifications. The issues may include incompatibility with
the eye trackers mouse emulation, too small targets, need
for text input, numerous keyboard controls, and action
sequences that require fast and precise pointing with a
mouse.
3.8 Level jumping (platform games)
The platform game genre involves horizontal movement
along platforms shown on the display and jumping from
Univ Access Inf Soc
123
one platform to another while avoiding obstacles and col-
lecting points or other items for later use. Originally, level
jumping games used a simple two-dimensional bitmap as
the background for the game animation. Later when three
dimensional graphics have become commercially viable,
platform games in three dimensions have been produced as
well. Three dimensional platform games are sometimes
difficult to distinguish from 3rd person action and adven-
ture genre. Consequently, the possibilities for eye control
are similar.
No eye controlled platform games are known to have
been developed. The basic controls are simple. One
degree of freedom plus a button for jumping is enough
for basic operation. However, eye control may not be
simple to implement. The playersmayneedtolook
around the display to be aware of dangers and to plan a
safe and effective route. Therefore, direct mapping of
gaze direction to movement may not be possible. Also,
careful timing of jumps may be challenging with eye
control. Controlling the game character in a three
dimensional world is more complicated, and many plat-
form games offer automatic help to the player. Still, it is
difficult to imagine how an efficient gaze controlled
version could be built. However, platform games are
usually single player games played without a network
connection. In the long run the player’s goal is to explore
new strategies and to beat his or her own record.
Therefore, even inefficient controls may afford rewarding
playing experience.
3.9 Turn-based strategy
Turn-based strategy games (e.g., Civilization) give the
player as much time as needed to complete one round of
information gathering and commanding of units in the
field. When all players have completed their turn, the game
simulates and displays the results and waits for the players
to react during their next turn. Assuming that the user
interface elements are big enough, turn-based strategy
games should be playable with eye controlled mouse
emulation. Unfortunately, this is not always the case
because of the same reasons mentioned for other turn-
based genres. The games often contain user interface ele-
ments that are too small to operate comfortably with an eye
tracker. Strategy and role playing games tend to have a
huge number of commands that the user can give at any
time. A field commander may have dozens of units to
command with many possible orders to give. When played
manually, the usual strategy is to select units with the
mouse, and give the orders through the keyboard. An
obvious alternative for eye controlled use is a menu.
Extended eye controlled use of menus can, however, be
distracting and tiresome. If the menus are too small for eye-
based selections, playing with the eye tracker is difficult or
impossible.
There is a special genre of simulators for simulating
sports teams, such as football (mostly the European
variant), or ice hockey teams. Some of these consist
mainly of large menu structures that display various
parameters of the team and the players. The games that
the simulated team plays are sometimes not simulated
in detail—only enough to present strategic options to
the player in order to make decisions on pulling away a
player, or pushing for victory even with the risk of
injuries. In other games the strategy game is combined
with a low level game simulator where the player con-
trols individual players during a game. In relation to eye
controlled play the strategy component is just like any
turn-based strategy game. The low level game simulation
component, on the other hand, is a mix of a real-time
strategy game and a 3rd person adventure where the
action occasionally escalates to situations similar to beat-
em-up games. Accordingly, the challenges for eye con-
trol are considerable as detailed in the sections on these
genres.
3.10 Turn-based role playing games
The eye control features of turn-based role playing games
are similar to other turn-based games. If sufficient time for
operating the user interface is available, the technology
used in the game does not interfere with the eye-mouse,
and the user interface elements are of sufficient size for
eye-mouse operation, the game should be playable out of
the box.
3.11 Real-time strategy
Real-time strategy (e.g., Command and Conquer by
Electronic Arts) is a game genre where the game does
not wait for the players to complete their actions.
Instead, events in the game world go on as the player is
sending commands to his units. Because of this the game
can overrun a player if he or she is slow because of the
use of eye tracker input. When playing against computer
opponents, the pace of the game can usually be adjusted.
Unfortunately, these games require very intensive mouse
usage. Therefore, players using eye control may have a
significant disadvantage when playing against human
opponents over the Internet.
3.12 Real-time role playing games
Real-time role playing games have become very popular
recently. Most notably, the World of Warcraft servers (by
Blizzard Entertainment) have attracted millions of players.
Univ Access Inf Soc
123
Eye control problems are similar to those encountered in
turn-based role playing games and three dimensional
adventure games. When playing against computer oppo-
nents the game may be interesting, but against human
opponents over the Internet, it may be impossible to design
eye control methods that would be competitive. However,
because of the cooperative nature of the game, it may be
possible to navigate the player’s character through the
game so that it can avoid fast-paced combat and still offer
enough support for its allies to earn its place in the social
hierarchy of the game world.
3.13 Racing
Racing games come in many forms. They range from very
simple games where the players task is to maintain full
speed and steer while hoping to avoid obstacles, to high
fidelity simulations. The point of view can be from within
the cockpit of whatever craft is used for the race, from the
outside, or even from a distance showing the whole track.
Some games allow the players to choose from several
camera positions. The control requirements are similar to
FPS games. Especially when the game world is viewed
from the driver’s seat, there is practically no difference in
the control requirements between a FPS game and a racing
game. The authors of this paper are not aware of examples
of successful eye controlled racing games.
Racing games also feature some dissociation between
the focus of attention and movement control. For example,
they often include gauges for speed, fuel, time, etc. Direct
mapping between gaze position and steering, for example,
is not easy because glancing at the speedometer would
cause steering actions. Overall, racing games range from
simple games where gaze-based control may be possible
given sufficient freedom to modify the game to games
where efficient gaze-based control is impossible.
4 Examples
The previous section has discussed the possibilities of eye
controlled games in general terms. Different game genres
have been described, outlining their position in the con-
tinuum of the level of effort that is needed to adapt the
games to eye tracker use. This section describes concrete
examples. It starts with a game that can be played using the
mouse emulation mode offered by most eye trackers that
are being offered for augmentative and alternative com-
munication (AAC) use, and it ends with a project which
involved a complete rewrite of the user interface. Apart
from the chicken shooting game (external modifications)
and the Eye Chess (complete rewrite) these examples have
not been tested or developed with players with disabilities.
4.1 Go over the internet using Cgoban3
Go is an ancient Chinese game played on a 19 919 grid,
that is usually drawn on an approximately half meter
square wooden board. Smaller 9 99 and 13 913 grids
are often used for quick games and for training beginners.
The gameplay consists of two players (black and white)
placing stones on the intersections of the grid. The objec-
tive of the game is to surround territories on the board. This
can be achieved by surrounding empty territory or
by capturing the opponent’s stones. Stones are captured
when one of the players manages to populate all intersec-
tions surrounding an opponent’s group of stones with his or
her own stones.
Go, like Chess, is a game of complex strategy and tac-
tics. Placing of stones is separated by contemplation that in
some cases can take hours or even days. In addition to
being turn-based, the pace of game is usually relaxed and
during a game the player needs to give only one kind of
command to place the stones. These properties make Go, in
principle, well suited for eye controlled playing. Therefore,
it was decided to try if it is possible to use Go software with
eye pointing. The test was performed using Cgoban3,
3
the
most popular western Go software. Some of the discussion
below is not Cgoban3 specific, but applies to Go software
in general.
The game window of Cgoban3 is shown in Fig. 1.As
we can see, most of the window is taken by the Go board.
In addition the right hand side of the window has a bar of
additional controls. At the top there is a number of buttons
for checking the game rules, resigning the game, asking for
the right to undo a move, etc. Next to these buttons is a list
of players and observers that are present in the game.
Below these are the player information and a chat window.
We used CGoban3 with a Tobii 1750 eye tracker that is
integrated into a 17 in. LCD display. The Windows control
functionality of the MyTobii software package was used
for mouse emulation. To facilitate eye tracker pointing, we
set the display resolution to 800 9600 pixels and
increased CGoban font size to 20. Cgoban3 scaled the user
interface elements according to the font size. The only
element that was difficult to use even with large font sizes
was the scroll bar that remained too narrow. We con-
strained ourselves to the 9 99 board. According to our
experience 13 913 boards might be playable, but 19 919
is beyond the upper limit of eye tracker accuracy, at least if
mouse emulation is used for pointing.
Gaze aware Go software could alleviate the problems
with the pointing accuracy. For example, we found the
dwell-time based mouse click emulation difficult to use
because the timer animation that MyTobii mouse
3
Cgoban3 is available for free at http://www.gokgs.com.
Univ Access Inf Soc
123
emulation displays leads to a drifting phenomenon that
complicates placing the stones. Namely, as the transparent
animation is displayed, the player’s eyes tend to follow it
rather than remain fixed on the position of the desired stone
placement. The tracker’s gaze point estimation is usually
somewhat off the mark. As the gaze shifts on the anima-
tion, it shifts further away from the desired stone position,
the eyes again follow it, etc. By the time the dwell-time
timer fires, the gaze may have drifted away from the
desired stone position, and the stone is placed on a wrong
intersection. A mistake like this can be fatal in a Go game.
If the pointing system could be made aware of the grid on
the Go board, the animation could be shown only on the
intersection that is nearest to the initial gaze position. This
way the iterative drift loop would not occur.
It turned out to be necessary to turn off the mouse click
emulation in between moves. Planning a move involves
lengthy periods of staring at various positions on the board
to mentally visualize the effects of moves and the follow-
ing move sequences. A dwell time based click emulation
would produce unwanted moves. The MyTobii Windows
control system has a fairly easy way of turning the click
emulation on and off, so this was not a problem.
Despite the difficulties, it was possible to play 9x9 go
with eyes only using Cgoban3. Unfortunately, in addition
to MyTobii mouse emulation, an on-screen keyboard for
text entry is needed. Text entry is necessary when logging
onto the server to enter user name and password. Text entry
is also desirable for chatting with the opponents and other
players. Players who do not respond to greetings and other
messages are often considered rude.
Apart from playing, online Go offers other activities.
The teaching functionality in Cgoban3 requires a lot of
menu usage and scrolling. Both of these activities are
troublesome with eye pointing. Observing teaching games
and following lessons given by others, on the other hand,
requires few interactions and is, therefore, fairly easy to do.
Observing the games of strong players and having access to
their games for learning purposes is one of the greatest
advantages that online Go offers to amateur players. Both
of these activities can be achieved with eye pointing on
CGoban3, although going through the games offline
requires a fair amount of clicking the buttons that control
the replay.
Overall, it appears that Cgoban3 could be used with an
eye tracker and mouse emulation. Playing on small boards
and observing games was easy. Using some of the more
advanced functionality and text based chat was more
difficult.
4.1.1 Chicken shoot (external modifications)
As an example of a game where external modification
enabled use with eye control, Joos et. al. (Dresden Uni-
versity of Technology) have augmented a two-dimensional
FPS game called ‘Chicken Shoot’.
4
This game is a classic
Fun Shooter with adventure elements. The main task in
Chicken Shoot is simply to blow the chickens away. In
each mission the player can collect a number of items and
receive extra points for various actions. The initial levels
are fairly easy, but the difficulty increases from mission to
mission. In normal playing mode the player can move the
cross-hair used for aiming with the mouse and shoot using
a mouse click. In addition, the 2-D scene can be navigated
by scrolling with the cursor keys on the keyboard. After 10
shots the player must reload the gun by pressing the right
mouse button. A screenshot of the Chicken Shoot game is
shown in Fig. 2.
In order to make this game playable solely through eye
control, an external eye gaze to mouse and-keyboard
wrapper was developed. The goal was to translate the eye
tracker data into mouse and keyboard events and then
inject these into the game. The most obvious and simplest
task was the positioning of the cross-hair, which was
achieved by directly driving the mouse cursor with gaze
position. The second task was to emulate the scrolling
functionality. This was achieved by using off screen gaze
aware regions on both sides of the screen. These regions
generated scroll events when the player’s gaze dwelled on
them longer than 500 ms.
Since the aim was to render the game controllable
through gaze alone, the trickiest task was to emulate the
shooting command. Since the content of the game is highly
dynamic, with both chickens moving in the scene, and the
scene itself also being navigable, a dwell time approach
was not feasible for two reasons;
Fig. 1 The game window of CGoban 3: White is taking a beating
from black who is using an eye tracker to play
4
http://www.chickenshoot.com.
Univ Access Inf Soc
123
(a) The natural types of eye movements in this dynamic
situation are smooth pursuit and saccades rather than
fixations and saccades. This means that one would
have to dynamically move the dwell time sensitive
regions according to the objects under scrutiny. Since
the 2 D object location is not accessible from outside
the game this approach doesn’t work.
(b) Even if one could compute dynamic dwell times for
smooth pursuit eye movements one would become
caught in the trap of the so called Midas Touch
problem [7], which means that either a lot of false
alarms would result in unintended ‘‘friendly fire like’’
shoots, or dwell time selection thresholds would need
to be set so high that the likelihood of hitting a
chicken would approach zero.
For these reasons, a switchable automatic machine gun
mode was implemented, using an off screen toggle region
above the screen. Once this machine gun mode was swit-
ched on, left mouse clicks were injected into the game at a
rate of 2 Hz until the user toggled the machine gun mode
off. Reloading of the gun was performed automatically by
counting left mouse clicks and issuing a right mouse click
after every 10th left click.
A restriction of the approach became obvious during
testing. Although testing was completed with the LC
Technologies EyeFollower system, sampling at 120 Hz,
the 20 sample delay introduced by the adopted gaze
smoothing algorithm complicated the cross- hair position-
ing control due to the temporal delay and associated spatial
offset. Reducing to 10 the number of gaze samples used for
smoothing solved this problem. Therefore, in order to
control this game, eye tracking devices must have a higher
sample rate and sufficiently low noise levels in their data.
The final system was tested by eight subjects at Dresden
University of Technology, as well as by participants of the
2006 COGAIN Camp in Turin, Italy. Results showed that
subjects’ performance with the eye controlled system was
initially worse than when using mouse and keyboard for
playing. However, after four to five trials, most subjects
outperformed the mouse and keyboard control condition in
the final score. This result can be explained by the faster
positioning of the cross-hair by gaze compared to the
manual positioning which involves both a relocation of
attention to the target and a subsequent manual relocation
of the positioning device.
The final judgment of the successful adaptation to eye
control is the subjective experience of the player. In this
case, players reported a strong feeling of immersion in the
game, and lower feeling of impediment in the eye con-
trolled mode than in the mouse and keyboard mode. For
those players that do not have the option of playing with a
mouse due to physical disabilities (all disabled participants
in the above-mentioned test in Dresden had a late stage
ALS diagnosis) a direct comparison cannot be made.
However, the physically disabled showed strong feelings of
joy being able to play a game after many years when this
had been impossible for them.
4.2 First person shooter (internal modifications)
Three projects have investigated the use of eye trackers in
first person shooter games. Jo
¨nsson finished her Master’s
thesis in 2004, Smith and Graham reported their results in
2006 and so did Isokoski and Martin. Smith and Graham do
not provide details on their implementation, but Jo
¨nsson as
well as Isokoski and Martin [6] and Isokoski et al. [7]
found that they needed to modify the game in order to
make eye tracker control possible. Jo
¨nsson modified a
game called Half Life, Smith and Graham worked with a
Java-based implementation of Quake 2 known as Jake2,
and Isokoski and Martin used a game-like FPS input device
testing framework originally implemented by Laurent
Gomilla and Maurice Svay. A screenshot of the display of
this testing platform in eye tracker mode is shown in Fig. 3.
In an FPS game the direction of the weapon and the field
of view are usually connected. There is a crosshair in the
middle of the display and aiming happens by rotating the
field of view, so that the target is under the crosshair. In
other words, in order to aim, the player must manipulate
the field of view so that the target is at the middle of the
display. Usually the field of view can be rotated by moving
the mouse or displacing a stick on a gamepad.
There are different ways that the eye tracker input could
be used in a FPS game. Jo
¨nsson implemented a mode
where one could aim the weapon with the gaze and move
the field of view with a mouse and a mode where the
weapon and the field of view followed the gaze. Isokoski
and Martin reported only the former, Isokoski et al.
Fig. 2 The Chicken Shoot game; the cross-hair slightly above the
center of the screen is controlled by gaze
Univ Access Inf Soc
123
reported both, and Smith and Graham reported only the
latter.
Jo
¨nsson did not report the performance of the partici-
pants in her experiment. Smith and Graham and Isokoski
and Martin found that their participants performed better
with manually operated pointing devices. Isokoski et al.
concentrated on various levels of distributing work
between eye tracker input and gamepad input, finding
similar performance regardless of the distribution.
Jo
¨nsson did not have enough participants to perform
statistical tests on the data that she collected on the par-
ticipants’ opinions and impressions regarding eye control
in FPS games. However, she reported that the participants
experienced the use of eye trackers in this context posi-
tively. Smith and Graham did not report statistically
significant differences in player performance between
mouse operated aiming and eye operated aiming. Their
participants reported higher level of immersion with eye
controlled aiming. In both cases, it is difficult to assess the
significance of these results. The participants had a very
short exposure to the eye tracker. With longer playing
sessions their skills and opinions might change. Isokoski
and Martin report longitudinal data, but only for one par-
ticipant (Isokoski himself). Isokoski et al. found differences
in the consumption of ammunition and shooting distance
when more work was done with the eye tracker input.
However, the longest experiment [5] lasted only for ten 5-
min sessions per input device configuration. Therefore, the
long-term effects of learning to the relative efficiency of
manual input devices and eye trackers remain unknown.
Efficiency is a central issue in FPS games, because they
are very popular in competitive gaming over the Internet.
Eye trackers are unlikely to be popular if using them leads
to losing the game.
Both Jo
¨nsson and Isokoski and Martin report the need to
alter the internals of the game software in order to make the
eye tracker work. The most important reason for this was
that they wanted to make the aiming crosshair follow the
gaze rather than staying fixed in the middle of the display.
Smith and Graham do not give details on their technical
implementation. They did not need to make the crosshair
move, so it is possible that they did not need to modify the
behavior of the game.
Jo
¨nsson used a fairly early tracker model from Tobii
technology, and possibly because of that decided to
implement a two-part software where a server process
outside the game communicated with the tracker and then
transmitted the eye gaze data to a module within the game
that controlled the game according to the gaze input. Iso-
koski and Martin were using a later model of the same
tracker. They used a game-like input device testing envi-
ronment where a dedicated input module was to be written
for all different input devices. Thus it was natural that one
was implemented for receiving the input from the Tobii
tracker as well.
Overall, it appears that some ways of using eye trackers
in FPS games require internal modifications of the game.
Note, however, that none of these examples suggests that
FPS games could be played efficiently using only eye
tracker input. In all of the known examples moving in the
game world and shooting still happened with manual input
devices. If these functions need to be gaze-operated as
well, it is even more likely that modifying the source code
of the game is necessary.
4.3 Eye chess (a complete UI rewrite)
Eye tracking seems to offer many opportunities for inter-
esting entirely new kinds of games that utilize the gaze
position information. However, no such games are known
to be available. So far, most games that have been written
with eye trackers in mind are adaptations of old games. In
some cases a game can be used directly, like Cgoban3. In
other cases small internal modifications are needed, as it
was the case with the FPS games. Sometimes, however, it
is better to re-write the whole user interface to better
support eye tracker usage. This is what Oleg Spakov did in
EyeChess [13]. A screenshot of the Eye Chess playing
window is shown in Fig. 4.
As was mentioned in the context of Go, dwell time
based mouse click emulation is troublesome. If no feed-
back on the dwell time timer is shown, accidental clicks
can lead to a lost game. If feedback is shown in a generic
Fig. 3 The display of the FPS input device testing system used by
Isokoski and Martin. Notice the two crosshairs. The big white one (on
the left edge of the leftmost penguin tablet) was the standard FPS
crosshair that stays in the middle of the screen. The small red
crosshair (on the neck of the penguin tablet on the right) followed the
gaze
Univ Access Inf Soc
123
pixel pointing mode where the pointer is not aware of the
squares of the game board, the feedback can lead to drifting
to neighboring squares. In EyeChess the squares of the
chessboard show the feedback on the selections. In addi-
tion, the squares have a center marker, so that it is easy to
focus the gaze on the center of a square in order to maxi-
mize eye pointing accuracy.
Interaction with the chess pieces happens by first staring
a piece to select it for moving and then staring at the square
where the piece is to be moved. In addition to the inter-
action with the board, games like Chess have a large
number of options to set. The user interface elements in
EyeChess have been made large enough to use easily with
eye pointing. The changes in comparison to mouse oper-
ated chess programs may not appear large, but they are
necessary. One critical control that is too difficult to
operate with eye control can render the whole game
unusable.
EyeChess is compatible with the MyTobii environment
and has been demonstrated for example at the COGAIN
demonstration day in Turin Italy. The reaction from chess
players, especially those with disabilities that hinder
manual play, has been very positive.
5 Conclusions
Interest in eye controlled gaming has increased. Most of
the research projects mentioned in this paper have taken
place during the last few years. It is expected that this wave
of interest will continue until academic researchers have
satisfied their initial curiosity. Further progress will depend
on whether commercial endeavors in the area are suc-
cessful or not. The effect of some critical factors is difficult
to anticipate. For example, a lot of computer gaming
happens in a living room setting. In addition to online
collaborators and opponents, friends that are playing or just
hanging around are often present. Remote eye trackers are
well suited to situations where the computer is used by only
one player. In multi-player situations eye control may be
more difficult to use, or require several head mounted eye
tracking devices.
For people with disabilities which interfere with manual
control in computer gaming, eye control offers an oppor-
tunity to play many games. Turn-based games and puzzle
games can usually be adapted for eye tracker use. Adapting
games that require constant and fine control is more diffi-
cult, and even when possible the adaptations tend to require
access to the source code of the game. However, these
same games offer the greatest opportunities to improve the
gaming experience of people without disabilities. Many
games require so much complex fine motor control that
potential players are overwhelmed with the arrays of sticks
and buttons they must operate. If some of the input can be
achieved effortlessly with an eye tracker, the gaming
experience may be improved.
While participants in the early studies on eye controlled
gaming have been very positive in their assessments of the
eye controlled modes in comparison to the traditional
manually controlled modes, there is still room for skepti-
cism. Eye tracking technology tends to evoke positive
reactions. When a person is using an eye tracker for the
first time with a well designed interface, the experience
may feel as if things happen merely by thinking about
them. This experience may distract the participants in short
studies from critically assessing their performance with the
eye tracker. In the long run, the initial enthusiasm may
wear out along with novelty, and without smooth game
play attitude may turn against eye control.
Eye controlled gaming is in its infancy. It is too early to
draw final conclusions on the value of eye trackers as
gaming devices. In the near future interesting demonstra-
tions, prototypes, and experimental results are expected. It
is time to assess the situation after seeing this evidence. So
far, eye controlled gaming seems promising for people with
and without disabilities.
References
1. Abrams, R.A., Meyer, D.E., Kornblum S.: Speed and accuracy of
saccadic eye movements: characteristics of impulse variability in
the oculomotor system. J. Exp. Psychol. Hum. Percept. Perform.
15(3), 529–543 (1989)
2. Dodge, R., Cline, T.S.: The angle velocity of eye movements.
Psychol. Rev. 8, 145–157 (1901)
3. Dorr, M., Bo
¨hme, M., Martinetz, T., Brath, E.: Gaze beats mouse:
a case study. In: Proceedings of COGAIN, pp. 16–19 (2007)
Fig. 4 EyeChess playing window. The square of the selected piece is
highlighted, and the square that the user is looking at is shown with
the green border
Univ Access Inf Soc
123
4. Duchowski, A.T.: Eye Tracking Methodology: Theory and
Practice. Springer, London (2003)
5. Isokoski, P., Hyrskykari, A., Kotkaluoto, S., Martin, B.: Gamepad
and eye tracker input in FPS games: data for the first 50 min. In:
Proceedings of COGAIN, pp. 10–15, (2007) available at
http://www.gogain.org
6. Isokoski, P., Martin, B.: Eye tracker input in first person shooter
games. In: Proceedings of COGAIN 2006: Gazing into the
Future, pp.78–81, available at http://www.cogain.org
7. Jacob, R.J.K.: The use of eye movements in human-computer
interaction techniques: what you look at is what you get. ACM.
Trans. Inf. Syst. 9, 152–169 (1991)
8. Jo
¨nsson, E., If looks could kill—an evaluation of eye tracking in
computer games. Master’s Thesis, Department of Numerical
Analysis and Computer Science, Royal Institute of Technology,
Stockholm, Sweden (2005)
9. Klochek, C., MacKenzie, I.S.: Performance measures of game
controllers in a three-dimensional environment. Proceedings of
Graphics Interface 2006, pp. 73–79. CIPS , Toronto (2006)
10. LC Technologies, Eyegaze User Manual, September 2006
11. Lin, C.-S., Huan, C.-C., Chan, C.-N., Yeh, M.-S., Chiu, C.-C.:
Design of a computer game using an eye-tracking device for
eye’s activity rehabilitation, optics and lasers in engineering,
42(1), 91–108, Elsevier (2004)
12. Smith, J.D., Graham, T.C.N.: Use of eye movements for video
game control.. In: ACM Advancements in Computer Entertain-
ment Technology (Hollywood, CA, USA, June 14–16, 2006).
ACE 2006. ACM Press, New York (2006)
13. Spakov, O.: MyTobii Applications available at: http://www.cs.
uta.fi/*oleg/mytobii.html. Accessed 28 Jan 2007
14. S
ˇpakov, O., Miniotas, D.: EyeChess: a tutorial for endgames with
gaze-controlled pieces. (abstract) In: Abstracts of ECEM13,
Bern, p. 32 (2005)
15. Starker, I., Bolt R.A.: A gaze-responsive self-disclosing display.
In: Proceedings of CHI 90, pp. 3–9. ACM Press (1990)
16. Wade, N.J., Tatler, B.W.: The Moving Tablet of the Eye: The
Origins of Modern Eye Movement Research. Oxford University
Press, NY (2005)
17. Wikipedia, Lines (video game) available at. http://en.wikipedia.
org/wiki/Lines_(video_game). Accessed 28 Jan 2007
Univ Access Inf Soc
123
