Acta
Psychologica
71 (1989)
217-242
217
North-Holland
TRAINING OF EFFICIENT OCULOMOTOR ENHANCES SKILL ACQUISITION Kimron
L. SHAPIRO
and Jane
STRATEGIES
E. RAYMOND
*
University of Calgary, Canada
The
present
acquired
study
through
investigated
practice
the
hypothesis
complex
visuomotor
different
set of simple tasks, or drills. One group,
received
training a second
group
training
designed
to increase
video
designed
with practice
to
training
and matched
training
exhibited the inefficient
significant
minimize
of subjects,
the frequency
to a control
for total
significantly or control
inverse correlation in the acquisition
movements
oculomotor
behaviours
and
Performance
the number
path
behaviours,
group,
training
received was inter-
groups
but receiving
The group receiving
efficient
to a
group,
in the
no specific oculomotor
in the video game and fewer foveations from each other.
of foveations
perceptual
motor
Overall
than
there was a
in the game and game score. The
in terms of their implications
of any complex
scan
be to a
were exposed
experimental
two experimental
the video game
can
subsequently
experimental Oculomotor
of these
group playing
which did not differ
between
optimize
of eye movements.
performance
groups,
eye movement
eye movement
time in the experiment. superior
results of this study are discussed training
eye
the efficient
the inefficient
on the video game.
game was compared
either
efficient
tasks and can be transferred
task, such as a video game. Each of two groups of subjects
whereas spersed
that
on a series of simple
for the importance
of oculomotor
task.
A video game is a predominantly visual task. As such, the cognitive and motor skills which determine game performance depend on an efficient flow of relevant visual information (Birch and Lefford 1967; Howard 1971; Williams 1974). Since eye movements act to place specific images on the fovea for detailed visual analysis, the perceptual, cognitive
and visuomotor
systems
critical
for game
play cannot
per-
form optimally unless the oculomotor system is operating at maximum efficiency. Eye movement control is particularly important when the * The authors the resources assistance draft
wish to thank necessary
Donchin
the present
in the initial stages of the project,
of the manuscript.
laboratory
We would
at the University
Requests Calgary,
Dr. Emanuel
to accomplish
2500
for reprints University
OOOI-6918/89/$3.50
should
Elsevier
Dr. Barry
and Dr. N. Moray like to thank to K.L.
Calgary,
Science
Shapiro,
Alberta,
T2N
Publishers
for helpful
in collecting Dept. lN4,
of Illinois
Dworkin
the personnel
for their assistance
be sent
Dr. N.W.,
0 1989,
also
of Calgary
and the University work,
for providing
for his advice
comments
and
on an earlier
of the Vision/Cognition and analyzing
of Psychology,
Canada.
B.V. (North-Holland)
the data.
University
of
218
K. L. Shaprro.
J. E. Rqwond
/ Trarnrng
oculomo~or
strategies
visual scene is dynamic in nature (e.g., Senders 1983) as is the case in a video game. The present study investigated the role of oculomotor control on the acquisition of video game skills. The experiment was designed to: (1) determine whether specific patterns of oculomotor behaviour could be trained, and (2) assess the influence of specific eye movement strategies during acquisition of skill in the project-wide video game, Space Fortress. (For a complete description of the game, rules, stimuli, and required responses, see Mane and Donchin 1989, this vol.) Since the visual, cognitive and motor tasks required in a video game are simplified versions of tasks found in a variety of sport, vehicle guidance, and other situations, the results of this study have implications for training skill acquisition in more complex environments. Two classes of eye movements are critically important for playing most video games and similar perceptual tasks. Saccadic eye movements, characterized by abrupt, rapid and ballistic eye movements, act to detect and correct for differences between target location and eye position, thus producing foveation of a target. The saccadic system is used both when searching a static visual display and when initiating foveation of moving objects. The second class of eye movements are smooth, slow movements which are typically stimulated by relatively slowly moving objects with smooth trajectories. These eye movements, known as smooth pursuit eye movements, are used to match eye velocity to target velocity for the purpose of tracking a moving object, i.e., maintaining foveation after an initial saccade to the object. Foveation during object movement permits the acquisition of important information about object velocity and shape. When viewing a dynamic display, the two eye movement systems are used sequentially to produce a scanpath. A scanpath can be defined as a series of saccadic eye movements alternated with short fixations (of stationary objects) or foveations (of smoothly moving objects) which allows the viewer to sample the visual environment discretely. Cognitive control over the scanpath permits an observer to choose various strategies for assessing the information in a display. For example, when the observer knows where to find important information, increasing the number and rate of saccades when viewing a stationary scene increases the opportunities for data sampling and thus increases the probability of appropriate selection of input information (Festinger 1971). In dynamic viewing situations, however, where objects may appear and disappear unexpectedly, it can be argued that there are disad-
K. L. Shapiro, J. E. Raymond
vantages to frequent sampling. area may decrease detectability
/ Training oculomo~or stra&gies
A saccade to an inappropriate of objects appearing at another
219
viewing location
of the display. Since saccadic eye movements are extremely fast, visual detection is reduced during the eye movement (Matin 1974). The more frequently saccades are made, the greater the proportion of time spent with a functionally blind visual system. Furthermore, the initiation and execution of a saccade may cause a drain on attentional or cognitive resources. Thus, efficient use of the oculomotor system should enhance performance in dynamic tasks requiring fast responses by controlling and limiting the flow of visual and cognitive information. Learning to perform a task optimally may therefore involve the acquisition of efficient, as opposed to inefficient eye movement or scanpath, strategies. Indeed, research on scan path patterns adopted in cockpit situations has shown that observers spontaneously develop efficient patterns of eye movements in complex tasks (Fitts et al. 1950; Clement et al. 1968; Senders et al. 1964). Senders (1983) has shown that observers learn to respond to the statistical structure of dynamic displays in a truly optimal manner. In the present paper, we investigated whether this process can be facilited with appropriately designed training regimes. Although placing the image of an object on the fovea for detailed processing is the primary function of the oculomotor system, images landing on peripheral areas of the visual field also undergo substantial visual processing. A number of investigators (e.g., Eriksen and Hoffman 1972; Posner 1980) have demonstrated that under specific conditions, a subject can dissociate the focus of attention and the point of fixation. Thus when an object appears in the periphery, a subject may utilize the albeit reduced visual processing capacities of the peripheral fields by directing attention to that area and eliminating the costly effort of making an eye movement to fixate the object (e.g., Senders et al. 1964; Moray 1984). If the object is large enough to be effectively analyzed by the peripheral areas of the visual field, then shifting or widening attention to peripheral areas may be an important contributor to an efficient oculomotor strategy, assuming that attentional shifts are less costly and faster than oculomotor shifts. Efficiency of visual search in a continually changing display may involve: (1) eliminating repetitive saccades to stimuli which had been previously analyzed (assuming memory for what had been previously viewed is perfect), (2) eliminating saccades to stimuli known to be
220
K. L. Shapiro. J. E. Raymond
/ Trmning oculomofor
.stmtegies
invariant and previously found to be irrelevant, (3) eliminating eye movements used to foveate images which can be analyzed sufficiently with the peripheral visual field, (4) eliminating reflexive tracking of irrelevant moving of the task.
objects,
and (5) matching
attention
to the demands
Two of the main goals of the present study were to: (I) identify the oculomotor strategies used by naive subjects as they acquired skill in and (2) document differences in the video game, Space Fortress, asymptotic eye movement behaviours between poor and skilled players. Based on pilot studies and other reports described below, we hypothesized that performance on the game would be improved if subjects exhibited an efficiency of eye movements as described previously. The design of the present experiment was to train one group of subjects to adopt an efficient strategy, train a second group to adopt an inefficient strategy, and then compare their performance in the video game with a temporally equated control group receiving no specific eye movement training. There have been several studies relating oculomotor efficiency with skilled performance in static viewing situations. Yarbus (1967) demonstrated that, when asked specific questions about certain objects in a picture, subjects concentrated their fixations on those objects and failed to saccade to irrelevant objects. This finding demonstrates that scan patterns are not random and that subjects exhibit cognitive control over them. Other researchers have demonstrated that when repeatedly asked to view the same picture, subjects adopt specific patterns of fixations suggesting a learning component in the acquisition of scanning strategies (Yarbus 1967; Jeannerod et al. 1968; Norton and Stark 1971). Considerable research on scan path behaviours in airplane cockpits has also shown that subjects develop specific eye movement patterns which reflect the meaning of symbols in the display as well as their statistical properties (Fitts et al. 1950; Clement et al. 1968; Senders et al. 1964). Kundel and Nadine (1978) also applied these ideas to skilled performance by examining differences in the scan paths of novice and expert radiologists viewing lung X-rays. They noted that experts make shorter fixations than novice radiologists. Similar results were reported by Schoonard et al. (1973) who investigated scan patterns of factory workers inspecting integrated circuit chips. Schoonard et al. (1973) reported that, in this specific task, accurate inspectors made fewer
K. L. Shapiro, J. E. Raymond
/ Training or&motor
saccades and shorter fixations. Data that scan path behaviour is learned.
strategtes
from all of these studies
221
suggest
Smooth pursuit eye movements also may be improved with practice. Bahill and LaRitz (1984) reported that expert baseball batters had faster smooth pursuit eye movements and better control over the head and eye coordination necessary for gaze stability than amateur batters. Moreover, Ludvigh and Miller (1954, 1958) reported that dynamic visual acuity, a task largely dependent on the ability to smoothly track a high speed target, improves with practice. The above studies suggest that practice can modify eye movements in a manner that is correlated with improved performance on a task and that oculomotor behaviours for a given task may be habitual in nature. However, in these studies changes in the eye movement behaviours have occurred through repeated exposure to the same task on which performance subsequently is measured. Thus it is not clear whether the oculomotor habits acquired through practice are simple by-products of learning other skills necessary to task performance, or whether acquisition of appropriate scan path behaviour is a more active process such as acquiring a specific manual skill. If the latter is the case, then scan path and general oculomotor behaviour may be generalizable from one task to another. The present study sought to examine in a direct fashion the role of learned oculomotor strategies on performance. If appropriate eye movement behaviour patterns, or habits, actually lead to better performance, then acquiring ‘good’ scan path habits in one task should improve performance on a second task, provided that the optimal eye movement strategy is similar for both tasks. Such an argument supposes that eye movement strategies are habitual and generalizable in nature and that training oculomotor behaviours may be an important aspect of training regimes for complex visual motor tasks. If, however, the acquisition of oculomotor habits are specific to a task and are not generalizable, then practice on one task should have limited effect on performance on a second task even if the optimal eye movement strategy is similar for both tasks. In the present study we devised two practice regimes, i.e., two sets of simple tasks (drills), which were designed to produce two different eye movement strategies: frequent or infrequent changes in foveation. Exposure to these drills is analogous to practicing piano scales in an effort to acquire the complex skills critical to playing an entire piece of
222
K. L. Shapm,
J. E. Raymond
/ Trarnmg oculomotor strategm
music. In other words, unless accurate control over finger movement is initially acquired, the cognitive skills required to read and interpret the written music cannot be manifest in performance. In the present context, practice on the ‘inefficient’ or ‘efficient’ drills should produce the adoption of corresponding inefficient of efficient oculomotor habits. If oculomotor habits are generalizable from one task to another and eye movement behaviour plays an important role in performance, then practice on the inefficient versus efficient drills should lead to decremental or incremental differences in performance on the video game, respectively.
Method Subjects A total SD = 1.3) subject
of 33 male undergraduates were paid
selection
or uncorrected the experiment
procedure according
(described
scores
during
ranging
in age from
of the study.
appears
visual acuity
were no significant task
participants
elsewhere
differences
chapter)
description
All subjects
The distribution
skill level of subjects
in score between
in a previous
17 to 22 years (M = 18.9;
complete
in this volume.
of at least 20/30.
to entry
A more
to the various
is described
groups
groups
in table
on the project-wide
or on the mean
of each
of the
had corrected
subject’s
of
1. There screening
five lowest
game play.
Appuratus The video game, Space and Donchin
Table
(1989,
Fortress,
this volume).
was played using the equipment Eye movements
were recorded
described
in Mane
using an Eye Trac 210
1
Assignment
of experimental
and control
Controls
(N)
groups by entry Efficient
level Inefficient
group
reg.
e.m.
e.m. [l. 121
e.m.
cm.
(2)
(6)
(9)
(7)
(9)
group
Mean entry screen
score
1,020
1,007
1,040
- 936
- 1,565
- 934
2,981
2,547
3.461
1,087
1,017
Mean low a game score
-1,125
- 1,098
Mean high a game score
a Low or high game score
is defined
as the average
3,412 of a SubJect’s
personal
2,591 worst or best scores.
K. L. Shapiro, J. E. Raymond infrared
eye movement
used to support
monitoring
the subject’s
device
recording
apparatus
surement
was also
recorded
fixation
were superimposed
and white
video camera)
video
tapes
which
appeared
different speed
About
tape.
Signals
(filmed
Calibration
prior
indicating
played.
white
crosshairs
eye movement one
quarter
of the
half
of the tapes
and about
group were scored by observers and also naive to the purpose
on a video
records
using
records
mea-
the location
of
with a Sony HVC 2800 black
who categorized
by observers
of the
to oculomotor
were viewed
scored
observers
on video
223
tape. A chin rest was
recordings.
of the game or drill being
as a moving
playback.
different
on video
was made immediately
on a video recording
subsequently
observers
and stored
head during eye movement
eye movement and
/ Traming oculomotor strategm
The
resulting
picture
video
were
of the game.
tape players
scored
randomly
from
Inter-observer
the game followed
the project-wide
twice
each
who were blind as to the group assignment
Five
with slow
at least
chosen
of the experiment.
composite
eye movements
by
subject
of a subject
reliability
was very high
(r = 0.96). Procedure All subjects outlined Illinois
control
had their groups eye
when playing
in a previous
movement day
movement
inefficient
according
all efficient
reading
instructions This
and inefficient
instructional subjects
on the third session with two blocks
had
their
appearing
All remaining
efficient
eye
routine
subjects
in table
Three
experimentals
movements
recorded
2. A fourth
and every
group
recorded
as
of the
of the experiment.
the game during
played
(eye
on the first
time eye movements
control
etc. Within
after viewing practice
Each
by a block
video
except with
of these session
again were recorded
followed
respect
group were performed
approximately
in time (in the experiment)
began
on each drill
consisted
of eight
for all subjects. [1,12] group.
the
of
to viewing on blocks
schedule
the
the instructional oculomotor
Eye
4, 6, 8, and
eye movement
the limits of practicality,
control
chapter),
was introduced
one block of each drill plus one block of
the efficient
group
tape
the rest of the
of practice session
of
for two
of all subjects
during a single block of the game on session
eye movement
groups.
instructional recordings
Drill
until the ninth session.
twelve comprised
identically
eye movement
session,
consisted
below)
the first quiz (see an earlier
of the game.
game play, followed
subjects,
movement
the quizzes,
In the second
of game play. The tenth and eleventh
9 in all experimental
in detail
the first project-wide
six blocks
and continued
which
groups
the first session
of game play. Eye movement
session.
also were recorded
eye
groups,
each drill (described
video tapes and writing
of game play. Session
movements
mental
group.
eye movement
experimental
by viewing
block
of (canonical)
and then four blocks
writing
control
two sessions
procedure
the exact
had their eye movements
for and practicing
was followed
all experimental
inefficient
schedule
experimentals)
were made during this initial project-wide
The
controls,
experimentals)
to the
and then one five minute
blocks
the regular on a least
standard
followed
day only.
For minutes.
recorded
(eye movement
(1,121 efficient
and twelfth
Two of the subjects
group and are termed
eye movements
of subjects
second
chapter.
efficient video
recordings
and tapes, of the
of game play that corresponded
to eye movement
recordings
for the experi-
group
Blocks played Cum. blocks Cum. minutes E.m. record time Drills
Ineffwient group
Blocks played Cum. blocks Cum. minutes E.m. record time Drills
Ef/iicient
Blocks played Cum. blocks Cum. minutes E.m. record time
Control group
1 1 26 26 I/
1 1 26 26 r/
5 5 40 20
1
Table 2 An outline of the procedure
7 8 61
7 8 61
7 12 75
2
for all groups
r/
4 12 87
I/
4 12 87
6 18 105 105
3
4 16 114 91 v
4 16 114 97 I/
8 26 145
4
r/
6 22 145
r/
6 22 145
6 32 175 175
5
6 28 166 155 I/
6 28 166 155 r/
8 40 215
6
6
r/
6 34 222
r/
6 34 222
46 245 245
7
6 40 258 232 I/
6 40 258 232 r/
8 54 2X5
8
6 46 294 294 r/
6 46 294 294 v
6 60 315 315
9
8 54 334
8 54 334
8 68 355
10
8 62 314
8 62 374
8 16 395 395
11
1 63 385 385 1/
1 63 385 385 I/
12
K. L. Shapiro, J. E. Raymond / Trainuzg oculomotor strategies All
experimental
project-wide
A block
of practice
to the subject
instructions
drill
group,
1 was designed
to discourage
objects
in the display.
moving
object
and direction
For both varied
as quickly
of the next
form of feedback just
object
beside
1.6 degrees
studies
On
suggested
subject
interval
over the appeared.
of the fire button
and then started
a
the speed
had no control
of the stimulus.
half of the trials
After
in an
a button
to move again
to the subject
the stimulus
trials,
Subjects
to the ship and to attempt to detect
between
contained
at
in the
of trials.
in the periphery.
that novices
appearing
saccades
the screen
and displayed
the remaining
of visual
to peripherally
(see a description
The dot was small enough
was necessary
(approximately
for a brief
appearing
The
RT was recorded
the ship.
unnecessary
For the inefficient
on each trial, although
after detection
on a random
fovea1 vision
vision.
foveation
Pilot
group,
(mine)
their
peripheral
trials.
were
Subjects
after the start of each trial, a stimulus
as possible trial.
Drills
and buttons.
and to promote
at the start of a trial,
between
at the end of a block
For the efficient
in the
each drill.
eye movements attention
at a fixed velocity
interval
the ship was immobilized
appearing
groups,
feedback.
the joystick
for use in ship control.
task was to press the fire button
the beginning
to direct
by the
Feedback
of drill 1 was to eliminate
attention
to encourage
At a random
chapter)
performance
including
the use of peripheral
of motion
The subject’s
to provide
the purpose
(ship) which drifted
ship’s movement.
moving
as required
in duration.
to read prior to playing
and train peripheral
objects,
press,
60 minutes
was two minutes
game equipment,
For the efficient
eye movements
previous
than
for drills 1 and 2 or a game score for drill 3 was displayed
at the end of each block
were given written
that
for each drill
time (RT)
using the regular
group,
no longer
drills
form of a reaction
Drill I:
lasted
standard.
Visuomotor
played
sessions
225
angle)
to detect
while
to detect
probably
was
a larger
in this group were instructed
(approximately
its presence
was a small dot
the stimulus
the mine with their
3 minutes the mine
its presence
need to fixate
of visual angle)
was large without
enough
foveation.
the ship in order
to learn
ship control. For the inefficient itself
group,
or in the other
appearance
of a small
order to detect second
small
appearance. button between
half,
dot
appeared
the mine
beside beside
mine and ship. Subjects
dot beside
required
it. The
subjects
the ship
at a random
on these trials the ship.
Thus
were instructed
was a mine by
possibility
to foveate
On trials in which a dot appeared
were instructed
a dot appeared
half of the trials the stimulus
with a small
dot beside
the dot’s presence.
Subjects
until
on a random a mine
beside
interval
to withold
of the
the mine
after
responding
they were encouraged
to keep their eyes searching
in
the mine, a the mine’s on the fire to saccade for mines
at all times. DriN 2:
For the efficient
saccades
to peripherally
group,
the purpose
appearing
objects,
of Drill
2 was to train subjects
to use peripheral
attention
to inhibit
when possible,
226
K. L. Shapiro, J. E. Raymond
and
to improve
location.
For
peripherally
eye
movement
the inefficient appearing
peripheral
elements
present.
button
on the screen
but rather
the subject
had to respond
required
responses
to be made
and displayed
that in addition
as possible
to the subject
a friend),
for a friend
a mine was also
and to press
half of the trials,
the ship, as in drill 1. When
the fire button
as quickly
to
for detecting
whether
the fire button
beside
by pressing
a fixed
saccades
to drill 1, except
for a foe. On a random
a dot appeared
with
used to indicate
a foe and a ‘3’ indicated
to press
button)
to objects to encourage
to use fovea1 attention
a small letter,
indicated
was required
(the IFF
not appear were
saccading
and to train subjects
or a foe (an ‘E’
The subject
different
when
drill 2 was to designed
of the game. This drill was similar
to the ship being present was a friend
accuracy
group,
objects
/ Trurning oculomotor stmtegirs
this occurred
as in drill 1. All trigger
and
an RT
in the form of feedback
a
a mine did presses
was recorded
for all
at the end of a block
of
trials. For
the efficient
screen. This
The letter
changing
group,
letter
the periphery.
provided
visual display (e.g.,
from
the letter
foveation
the fortress was small
to foveate
not present
unless
ship, however, needed
landmark DriN 3: ship
(without
(requiring
drill,
tracking
to execute
eye
movement
ship
drill 3 was designed the drill was designed were in control
was established began a mine
of motion before
drill
Drill
during
to encourage of the ship’s
response 1 and 2
the ship appeared
of a dot appearing a minimal
RT.
that good players phase
foveation saccades
beside
the
ship and mine No
stationary
tend to foveate to train
of learning.
For
the
efficient
or
the efficient
of the ship while, for the inefficient between
orientation to apply
and remained and
between
the was
drill.
ship and mine.
in space thrust).
The direction
constant
immediately
In drill 3,
but had no control
on the sceen. At a fixed interval location
or foe. The letter
3 was designed
this
(i.e., they are unable
in a random
task with
suggested
to encourage
when the ship appeared appeared
its
then the letter
a motor
of the mine requiring
it was friend
The possibility
control.
patterns
group,
the direction
that
Rather,
near the ship in drills
in the center
whether
in this particular
pilot studies
learning
group,
all subjects
angle)
all the time,
an eye movement), before
in
to the fortress.
of visual
it was present
of the ship so that saccades
the inefficient for subjects
As stated earlier,
inefficient
appeared
to determine
the mine was present.
encouraged
while
and then back
of the letter)
at the
are triangular
of a mine in the periphery.
eliciting
This
eye movement
that experts
often
in
in the
a map of the display.
0.5 degree
of a dot appearing
the letter
the mine
was provided
steadily
a landmark
of a specific
patterns
Since
foveation
above.
when a mine appeared
we determined
fixation
the presence
of the
as described
of the ship.
For the inefficient subject
studies,
Their
(approximately
be made. The high probability foveation
to developing
discrimination.
first had to be detected
encouraged
were
enough
alone could not have signaled
the mine could
In pilot
in the center
it provided
to the ship, to the mine,
for letter
had to be discriminated
except
was added because
as their landmark.
was needed
presence
information
intervals
the size and direction
the ship to the letter).
going from
continuously
at random
quite critical
in calibrating
game use the fortress Second,
letter
which is probably
may be useful
was displayed
but changed irrelevant
The stationary
map
nature
the letter
was stationary
through
over
of motion
a trial. A trial
after the ship appeared,
began
to track
the
ship.
K. L. Shapiro, J. E. Raymond / Training or&motor Subjects
in the efficient
quickly
as possible
group were instructed
by correctly
orienting
they were instructed
Additionally,
ship thus encouraging
to attempt the ship
to score a hit on the mine as
and
depressing
to shoot at the mine without
them both to detect
22-l
strategies
the
fire button.
taking their eyes off the
and shoot at the mine using their peripheral
vision. For the inefficient appearance
group,
a small dot appeared
the dot appeared
to begin hitting
was encouraged
to promote
rapid
a score which reflected Eye movement
into
periods. absent block
record
separately
the subject’s
analyzed
their proficiency
goal
A no-mine
of eye movements in each
a series
was defined
during
was different.
reduced
in the form of
which
as a block
a mine was present
and was terminated
period
both drill and game play mine present
Thus
distinguished
each
vs. absent,
game
between
mine
of time during
which
goal was to fire at the fortress.
A no-mine
block
and
the mine was
A mine period was defined
and the subject’s
was
no-mine as a
task was to hit the
began when a mine disappeared
from the screen
when the next mine appeared.
of foveations
The
primary
objects
function
in a visual
perspective,
of eye movements
display
it is of interest
time
period,
movement
movement
recording
measure
3 foveations also recorded
fovea.
how many
respectively.
with a fixation
because
mine and no-mine
processing
were foveated
or a smooth
in a
pursuit
eye
extracted
of foveations
of different
objects
made
For example,
if a subject
began
an eye
on
was the mean number
of the many
one of the measures
of the fortress,
object
image
information
objects
fixation
and then saccaded the
which an
different
Therefore
was the number
periods,
From
a steady
the foveation.
the ship for a short period
constitute
is to control
the
of whether
records
period
on
to know
was used to achieve
mine and no-mine
tracked
lands
regardless
from the eye movement
during
prematurely
drill.
game situations:
which
for whatever
reason
obtained
situation
of periods
period
of time during
during
Firing
to wait until
the ship and mine
in drill 3 was provided
with this particular
mine with a missile.
given
between
of the dot.
for the two distinct
and the subject’s
Number
after the mine’s
measures
The continuous was analyzed
detection
for all subjects
interval
were instructed
the mine. Thus saccading
their score on the drill. Feedback
since
at a random
either near the ship or near the mine. Subjects
the
back fovea
then saccaded
to the ship,
to the fortress, changed
this would
twice.
of times each game object
A related
was foveated
periods.
Previous object viewed A second the subject (no-mine
measure
periods)
percentage
made
from the eye movement
immediately
or a mine
viewed a given object the subject
extracted
was viewing
(mine
prior
periods).
prior to a successful
a successful
of times the subject
hit for each
records
to making The
concerned
a successful
total
number
which object
hit on the fortress of times
the subject
hit was divided by the total number type. This
viewed a particular
ratio
object
served
of times
as an index
while making
of the
a successful
hit.
228
K.L. Shapiro, J. E. Raymond
Proportion
strategies
of foveations
A third foveations
measure directed
dividing
/ Truimng oculomotor
calculated
from
to a particular
the number
of times
the eye movement
object
during
a particular
records
each period
object
was the proportion
of
type. This was derived
was viewed
by the total
by
number
of
foveations.
Results Game score analysis Since
performance
on the Space
game score, comparison will be discussed game scores
defined
value
subject, significant
differences
control
groups
groups.)
The
sures variable action groups
effect
were reached
(F(2,
tended
For this reason,
to the mean
measures
revealed
(F(1,
lowest
scores
described
data
from
the
hoc
tests
two
for each subject. for each
there
were
no
data from the two
efficient
effect,
among
recorded
a group,
previously,
groups
Post
scores
scores
by
on the game was
achieved
within
30) = 1038.67,
p -C 0.05).
represented
on this measure
to show high variability
(Since,
a non-significant
significance in their
ANOVA. as were
best
peak performance
of the five lowest
the sub-groups
combined
30) = 3.62,
did not differ
is probably
as the mean of the five highest
between
analysis
game
in each of the three groups
in all groups
blocks.
was compared
in a repeated
Fortress
all subjects
Subjects
on successive
operationally This
first.
among
experimental
but the repeated
mea-
p < O.Ol), as well as the interrevealed
on the game
(see
that while table
l),
the three
the efficient
4000 3000
efficient
inefficient
control
Group Fig. 1. The mean low and high game score for each group. as the average
of each indicated
subject’s
five personal
worst
Low and high game score are defined
or best scores.
respectively.
by the solid black bars and high score by the striped bars.
Low
score
is
229
K. L. Shapiro, J. E. Raymond / Training oculomotor strategws
. efficient 0 inefficient 0 Calgary control n Illinois control -2000
t
,
.
60
.
in experiment
(
,
300
240
180
120
time
.
360
(minutes)
Fig. 2. Group mean game score obtained on the last block of each session plotted as a function of time in the experiment (including drill practice).
experimental higher
group had a mean highest
than that of the control
the inefficient
experimental
score (mean = 3439)
group (mean = 2655;
group
(mean = 2590;
highest
score of the inefficient
experimental
control
group.
these data.
Fig. 1 illustrates
Performance
in the project-wide
t(13)
group
game,
t(22)
Space
approximately
= 2.23,
spent
sessions
in which
second both
eye movements
efficient
groups
control
minutes
into
i.e., playing
were recorded
represents
improved
the game
group.
and statistically
Using
control
the groups
consistently
probe
non-significant
drills.
the score
the same rate and marginally
At this point,
highly variable
as a function
or practicing
For
on the
As can be seen in this figure, performance
at about
group demonstrating
experimental
a single,
that of the
in fig. 2 which
plotted
on the last block,
group but at the same rate as the Illinois the experiment.
experimental
and inefficient smaller
of the screen,
last block was used in the analysis.
experimental
the Calgary 250
in front
from
is illustrated
shows the group mean game score of the last block of each session of time
The average
p < 0.05).
was not different
Fortress,
800 points
p -C 0.05) and that of
= 1.96,
only
the last block
difference
between
to differ
scores
with
session
of the subjects groups
the
than the control
of each
into the performance
of than
group until about
begin
higher
faster
which shows a
at the end of the
experiment. Eye movement
analysis
The mean number as a function 3. Total groups
of foveations
per no-mine
of time in the experiment
time in the experiment at each recording a significant
effect
= 5.10,
p -C 0.01)
no-mine
data over all sessions
and
was roughly
session
revealed
no
(see table
of group (F(2, significant showed
and mine periods
of the game plotted
is shown in panels A and B, respectively, equivalent
for experimental
2). An ANOVA 18) = 8.47,
interaction no significant
on the mine period
p < 0.01)
effect. effect
of fig.
and control
A
and session
similar
of groups
analysis
(F(4, of
data 72) the
but a significant
230
K. L.
A
efficient
A
n inefficient
0
control
<
m
20
100
165
approximate
245
time
A
20
165
100
approximate
300
in experiment efficient
245
time
390
(minutes)
390
300
in experiment
(minutes)
Fig. 3. The group mean number of foveations during the game plotted as a function of eye movement recording session for the no-mine periods of the game (panel A) and the mine periods of a game (panel
B). Total
time in the experiment
control
session
effect
(F(4,
ever, repeating effect
effect.
For both
experimental inefficient
72) = 3.56,
the analysis
groups
(F(2, group
had
significantly.
and inefficient
no significant
periods,
foveations
post
groups
had means
hoc tests
(p < 0.05)
By the last block per mine period
for experimental
interaction
from the last three sessions
group and that the control
a mean of 2.23 foveations control
fewer
and
equivalent
and
session (see table 2).
p < 0.05) with a non-significant
and no-mine
experimental
did not differ
p < 0.01)
on data
18) = 4.59, mine
was roughly
groups at each recording
than
sessions revealed
of the experiment,
and interaction group
experimental the efficient
per no-mine
of 3.16 and 3.54 foveations,
How-
a significant
that the efficient
the control
and inefficient
and 2.28 foveations
effect.
revealed
or the groups
group had period.
The
respectively.
for
K. L. Shapiro, J. E. Raymond / Training oculomotor strategies Table 3 Mean number
of foveations
in the drills.
Efficient group Inefficient group
the mine
periods
possible
experimental
groups
movement number
The number
highly
correlated
The
1.38 3.92
and 3.64
For
all three
over time, although
of foveations = 0.42,
period,
in the initial
groups,
there
this effect
experimental
prior
was
in each
of foveations
periods
of the drills
found
impact
for the efficient
and inefficient
the group mean for each drill showed than the inefficient
p < 0.001).
between
in the game (last session)
was significantly
and drill 3 (r(14)
p < 0.001)
= 0.70,
to minimize
obtain
relevant
category,
saccades
the mean
for each subject.
effect
18) = 3.86,
(F(2,
0.01)
foveations
and inefficient
experiment,
the efficient
0.97,
while
in the efficient
or stimuli
the mine
group
the inefficient
a stimulus
was foveated
effect.
effect
the control
(F(4,
or the inefficient
had a mean of 0.59 and control
period
was group
72) = 8.64, group
group
p -e
had fewer
( p < 0.05)
By the last block
foveations
groups
to
of the latter
a significant
the efficient
significantly.
foveation
in a mine
of session
As before,
did not differ
group
p < 0.01)
experimental
not requiring
on these data revealed
a significant
interaction group
of foveations
= 0.70,
and
of the
of the mine per mine
had mean
values
of 0.96
and
of times
the
respectively.
Related subject Fig.
stimuli
An ANOVA
of the mine than either
the control period,
of times
p < 0.05)
and a non-significant
for drill 1 (r(14)
Since the mine may be considered
number
determined
made
and drill 3
in drill 2 was non-signifi-
and the number
was to train subjects
to irrelevant
information.
group
p < 0.001)
but not for drill 2.
One of the goals of the experiment group
the groups
correlated
experimental
is shown in table 3.
= 9.65,
in the drills (last session)
to
behaviour.
that the efficient
group on drill 1 (t(8)
The difference
of foveations
in an effort
on oculomotor
comparing
= 8.77,
in
(although
of the game were
was also analyzed
the desired
T-tests
cant. The number
decline
and less so for the inefficient
for each of the three drills on the last day of the experiment
(t(8)
to the first eye
was more pronounced
group
A
is that the
a general
groups
fewer foveations
respectively.
sessions
p < 0.05).
of foveations
number
groups
in the mine and no-mine
if the drills did in fact produce
mean
per no-mine
to their drills immediately
exposure
session.
(r(29)
number
foveations
received
so) for the efficient
group. The
Drill 3
1.63 1.99
among
of gaze changes
determine
Drill 2
for the difference
recording
not statistically
Drill 1 0.92 3.24
3.12
and
explanation
231
to this
foveated
4 illustrates
function
finding
was a significant
the number
of foveations
of time in the experiment.
group
effect
8.50,
p < 0.01)
inefficient
difference
the region of the letter (indicating
(F(2,
18) = 4.91,
group
had significantly
to the letter
An ANOVA
p < 0.02)
and a non-significant
the mine was friend or foe). region
on the data
a significant
interaction more
in the number
whether
effect.
foveations
effect
per mine revealed of
session
period (F(4,
Post hoc tests showed
than
the efficient
as a
a significant
group
72) = that the
(t(77)
=
232
K. L. Shaprro. J. E. Rqwnond
/ Training oculomotor strategies
100
20
165
approximate Fig. 4. Group
mean number
or foe plotted efficient
of foveations
as a function
group
of time
are represented
time
245
300
in experiment
390
(minutes)
in the region of the letter identifying
in the experiment
by the closed
(including
triangles,
the mine as friend
drill practice).
Data
for the
group by the closed
for the inefficient
squares and for the control group by the open squares.
4.52,
Differences
p < 0.01).
between
By the end of the experiment, as many times
foveations
the subject
as the efficient
foveated
may also be considered digit
is only
approaches checking
relevant ten,
revealed
tests revealed ferences control
the
groups
analyses
to these objects foveated these
divided
groups bonus
the digit
and the information foveations For
periods
would
in this
Repeated
contribute
An ANOVA
effect
effect.
Post
half as many foveations = 1.98,
(mean = 0.26)
to
on these data
a significant
interaction t(14)
of hits
has accumulated
is required.
logged
p < 0.05)
groups
the control objects
to other
objects)
of total
subject
of hoc
of the
p < 0.05).
and
between
Since
effect
there
for all subjects (panel
than (i.e.,
the efficient the number
showed
Difthe
initially
Therefore,
the
an object
was
A comparison difference game
fortress
differences,
and are presented
it is
foveations
group.
for certain
were no group
A), subjects
more
was examined.
of cumulative
groups
However,
of times
no significant
of session
the number
experimental group.
make proportionally
foveations)
groups
(p < 0.001)
and inefficient than the efficient
groups
was foveated
displaying
box).
were averaged
the no-mine
that
and inefficient
an object
and a significant display,
memory.
fortress
The information
the subject
(mean = 0.33;
control
of accumulated
response
18) = 500,
group
to particular
the three
motor
one and a half the number
were non-significant.
by the number
for
hits
and a non-significant
and
shows
(as opposed
of times
values
(F(2.
were non-significant.
situation,
efficiency.
of fortress
group had approximately
efficient
that the control
proportion
effect
p < 0.05)
make more eye movements not clear
the number
it may improve
as the inefficient
and inefficient previous
In the no-mine
a different
although group
groups
had approximately
only a few hits have been
that the efficient
between
The
point
when
72) = 2.75,
digit (mean = 0.16)
group.
the number
at which
inefficiency,
(F(4,
group
as an index of eye movement
a significant
session
and efficient
the digit displaying
when
of the number
oculomotor
control
the inefficient
in panels
viewed either
elements hits,
of
between (i.e.,
the letter,
the proportion
of
A and B of fig. 5. the fortress
or ship
K. L. Shapiro, J. E. Raymond
233
/ Training oculomoior strategies
no-mine periods
End
Beginning
B
mine periods
8% 28
38X 15% 43x
beginning
end
Fig. 5. Mean percentage of foveations directed to objects displayed on the screen for all subjects in the experiment. Panel A shows the distribution of foveations for the no-mine periods of the game at the beginning and end of the experiment. The distribution for the mine periods of the game at the beginning and end of the experiment are shown in panel B.
on 89 percent displaying
of foveations
cumulative
the experiment,
the percent
more than doubled
with only four and one percent
fortress
hits and the bonus
of foveations
directed
(9% and 5%, respectively)
indicator,
directed respectively.
toward
toward the digit and bonus
at only a modest
the digit
By the end of indicator
cost to foveations
of the
ship and fortress. foveations (4%).
In the mine periods
toward
the letter
(8%).
By the end of the experiment,
objects
had at least doubled
9%). The only object 2.99;
p < 0.07)
block
was the letter
of the experiment, of their
foveations
(s.d.
= 0.08)
and
proportion
the
of foveations
with game score (r(20) A related looking
question
during
15%;
a group which
indicator
= 0.06)
control
group
on
a mine
during
= 0.06)
box,
mine periods
viewed of
=
By the last on
it on 12%
occasions.
and significantly
The
correlated
p < 0.01). whether
aspects
there were differences
of the game.
To
no-mine 80
these
(F(2.19)
or foe.
group
(s.d.
toward
significance
the letter
10%
few box
and information
as friend
was positively
relatively
directed
3%;
the inefficient
only
directed
and the information
approached
viewed
while
made to the letter = 0.63,
indicator,
difference group
initially (1%)
of foveations
bonus
designated
the efficient (s.d.
concerns
specific
B), subjects
the proportion
(letter,
for which
20%
(panel
the bonus
answer
in where the groups
this
question
were
the object
of
periods
T
viewing ship
viewing fort
mine
mixed
periods
60 50 40 30 20 10 0
Fig.
6. The
movement periods)
percentage styles
of subjects
described
or mine (mine
data for the inefficient
viewing ship in each
in the text just
periods).
Data
viewing fortress group prior
mixed
using
each
to a successful
for the efficient
group
of
the
three
categories
hit on the fortress
are represented
group by the striped bars and data for the control
of eye (no-mine
by the white bars,
group by the black bars.
K.L.. Shop&, foveation found
immediately
that subjects, tended
60%
prior
mine periods
to adopt
of
foveations
made
periods)
or mine
mediately
prior
(i.e.&p
tion of subjects similar
styles.
situation
three styles of oculomotor group
firing
a missile.
The mean
related
groups
in the choice
The
last
Using (r(29)
The
positively
of
and game
the statistical
obtained game
of of
to view the ship
between
periods.
the other
of choice
periods,
two
immediately
subjects
was not
among
half the subjects viewed
the
in the prior
to
the fortress
primarily
than the mean
foveating
for subjects
the
evenly
and was higher (by 20 hits). but foveating
the ship.
immediately the efficient
than either
analysis
prior
Differences
to a hit may be
group had an average
the inefficient
p < 0.01)
number
of
of the data
on the last session
score
and
or control
foveations
with RT to hit a mine (r(29)
made
recordings
of
groups
= 0.56,
of eye
of game performance.
were made, a significant of foveations
periods
in the
and the number
is the correlation
with measures
the number
and in the no-mine
of hits on the fortress = 0.60,
was
had changed
score
for which eye movement
between
= -0.49,
found. number
aspect
all subjects
chose
homogeneity
primarily
foveated
imboth
p < 0.05).
measures
correlation
between
(71%) in the no-mine
hits for subjects
that on the last session
of eye movements
movement
of object
fort
of the experiment
of foveations
ship and fortress
for subjects
evenly
group
the control
p < 0.01)
70 more hits on the fortress
28) = 4.31,
Correlation
= 3.72,
between that
to the observation
approximately
(r(29)
so, than
pattern half
the made
were fairly evenly distributed
of fortress
(t(15)
their foveations
not significantly
(F(2,
number
was 87 hits higher
viewing
to view the mine predominantly,
- in the no-mine
periods,
(2)
on the first and last session
this relative
a consistent
In the mine
periods.
that, while the distribu-
distributed
chose
where subjects
behaviour
did not adopt
predominantly.
between
group
a hit in the mine periods,
in the no-mine
distributing
were evenly
67% of the inefficient
no-mine
using each of the three categories
periods
in the efficient
in this group
It was the ship
the ship on at least
this distribution
and the fortress
to making
fortress
of subjects
for the mine and no-mine
subjects
control
It was found
at the beginning
by the end of the experiment,
either
of foveations
foveations
of subjects
Although
found
styles
a hit, 60%
(57%) in the mine periods
remaining
prior
of these
each
The majority
predominantly
their
or ship and mine).
styles
the experiment.
to making
among
during
(1) viewing
on at least
a hit, or (3) splitting
Fig. 6 shows the number
eye movement
The
prior
periods)
and fortress
for all groups,
substantially.
(mine
was examined.
to foveate
and the ship or the fortress
immediately
to making
tended
styles:
235
strutegies
or a mine
to fire a missile,
one of three different
(no-mine objects
/ Training x&motor
to a hit on the fortress
when preparing
or the mine during Subjects
J.E. Raymond
mine
p < 0.01).
(r(29)
= -0.39,
periods
also
For the no-mine
of foveations
negative
in the mine
periods
p < 0.05) was
situation,
was correlated
was
correlated the
negatively
p -=c0.001).
Discussion The present study investigated the hypothesis that efficient oculomotor behaviours: (1) are a fundamental requirement for successful per-
236
K.L.
Shapiro.
J.E.
Raymond
/
Trawmg
oculomotor
strategies
formance on a complex visuomotor task, (2) can be learned in a part-task context and, (3) can be transferred to a complex task. In order to examine this hypothesis, we exposed each of two groups of subjects to a different set of simple part-tasks (drills), so that we might train two distinct oculomotor strategies and examine their influence on the canonical video game, Space Fortress. One group, the efficient eye movement group, received training designed to: (1) minimize eye movements by eliminating repetitive saccades to stimuli previously processed and/or previously found to be irrelevant, and (2) encourage the use of peripheral vision thereby eliminating saccades to objects able to be analyzed sufficiently with the peripheral visual field. The other group, the inefficient eye movement group, received training designed to increase the frequency of eye movements and to encourage the foveation of all objects on the screen. While novice players tend to use the second of these two strategies, experienced players tend to use the former. Performance of these two experimental groups in the video game was compared to a control group receiving no specific training but matched for total time in the experiment.
Performance
in the game
Perhaps the most important conclusion to draw from the results of this experiment concerns the influence of the two distinct oculomotor drills on canonical game score. The group receiving oculomotor training for the purpose of promoting efficient eye movement behaviours exhibited a significant 30 percent higher asymptotic game score then either the inefficient or control groups, which did not differ from each other. This implies that the part-task training experienced by the efficient group was effective in modifying behaviour during video game play. To determine if our specific drill regimen caused the improvement noted in the efficient group as opposed to some non-specific benefit accruing from part-task training in general, we compared game score performance between the efficient and inefficient experimental groups. This is an appropriate comparison, since they were treated identically in all respects other than the eye movement strategy imposed by the drills. Specifically, part-task training for both groups included: (1) exposure to the same visual elements of the game, (2) similar cognitive requirements, (3) knowledge of results presented with the same
K. L. Shapiro, J. E. Raymond
237
/ Trainrng oculomotor strategies
frequency, and (4) equivalent practice with identical motor responses. Since the efficient and inefficient groups differed in asymptotic game score, whereas the inefficient and the control groups did not, part-task training on the game as a whole cannot account for the improvement witnessed in the efficient group. Rather, the enhanced performance of the efficient group must be due at least in large part to the specific nature of the eye movement training in the drills with little or no contribution from other aspects of the part-task training. While we expected some decrement in performance of the inefficient group relative to the control group, the lack of a difference between these groups could be due to the possibility that the drills for the inefficient group were not effective or that the inefficient experimental subjects simply did not transfer the trained oculomotor behaviour to the game situation. Support for the latter possibility can be found in the acquisition of eye movement patterns and is discussed in more detail below. An alternative explanation is that the subjects may have matched their visual attention to the statistical structure of the task in the whole game without any conscious decision to do so (Fitts et al. 1950; Senders et al. 1964) and thus overwhelmed the inefficiency training. Two additional points concerning the game score data can be made. First, the control group received as many eye movement recording sessions as the two experimental groups. Since, our local (Calgary) control group did not differ from the Illinois control group, the repeated eye movement recordings probably did not interfere with game skill acquisition. This allows the data from our two experimental groups to be compared to the program-wide control run at the University of Illinois, as well as allows inferences to be drawn concerning our findings
and
those
of
other
investigators
involved
in
the
project.
Second, there is a point to be made concerning the game score acquisition data from both of the experimental groups as well as both our control and the Illinois control groups. Such a comparison indicates that the oculomotor drills were not effective at speeding acquisition of game scores early in the experiment as would be expected, but were more effective in the last third of the experiment. Oculomotor
strategies
An examination of eye movement behaviours in the oculomotor drills themselves revealed that the drills generally were effective in
238
establishing
K. L. Shapiro,
J. E. Raymond
eye movement
/
Trainrng
behaviours
oculomotor
consistent
strategies
with their intent,
i.e.,
the efficient drills produced fewer foveations than the inefficient drills. This was true, however, only for drills 1 and 3 indicating that drill 2 may not have been designed appropriately. A likely explanation for this is that a minimum of two foveations were required ideally to perform drill 2 for both the efficient and inefficient groups. The other two drills required at least one more eye movement of the inefficient group than of the efficient group to perform the task ideally. The pattern of efficient eye movement behaviours which were trained by the oculomotor drills appeared to have transferred to the complex perceptual motor task requirements of the Space Fortress game. Such a conclusion is drawn on the observation that the efficient group had significantly fewer foveations in the game than either the inefficient or control group. That the efficiency of eye movements displayed by the efficient group may have contributed to their better asymptotic game scores is suggested by the negative correlation (which, of course, must be limited and is task specific) between the frequency of eye movements in the game and game score. Moreover, since there was a strong positive correlation between the number of foveations made in these drills and the number of foveations made in the game, it appears that the oculomotor habits generated by drills 1 and 3 were generalized to the game situation. These data are consistent with the Schoonard et al. (1973) report that good vs. poor performance of factory workers inspecting integrated circuits was associated with scanpaths containing fewer, as opposed to more, fixations. It is important to note in the present study, as well as in the Schoonard et al. (1973) study, that optimizing performance by limiting eye movements is task specific. The present data are, however, also consistent with the general notion that eye movement patterns necessary for good performance in a dynamic task reflect task dynamics and semantics. The effect of oculomotor training on the global measure of game score may have been mediated through an impact on the efficiency of specific aspects of the game, such as mean RT to hit a mine and the average number of hits made on the fortress. If initiating and executing an eye movement occupies a visual/cognitive resource, then the time required to react to a novel object (e.g., a mine) should be correlated with oculomotor efficiency. The positive correlation between the number of foveations and RT to hit a mine supports this claim. In the no-mine situation of the game, good performance entails maximizing
K.L. Shapiro, J. E. Raymond
the number
of hits on the fortress
/ Training oculomotor strategies
before
the appearance
239
of the next
mine. Again, an extra eye movement might be expected to drain visual/cognitive resources necessary for the task of hitting the fort. The negative correlation between the number of foveations and the number of hits on the fortress supports the argument that oculomotor efficiency is related to cognitive and visuomotor efficiency. Not only was the goal of the drills to minimize foveations but also to discourage repetitive saccades to objects which had been processed previously. In the mine periods of the game, a single saccade to the letter indicating whether the mine was friend or foe is sufficient ideally. In actuality, by the last two sessions of the experiment, the efficient group made an average of one saccade to the letter. The inefficient group. on the other hand, made approximately 1.5 saccades suggesting an oculomotor inefficiency. This finding could be interpreted simply to reflect the overall group difference in the number of foveations discussed earlier. However, an analysis of the proportion of all foveations that were directed to the letter reveals that the distribution of oculomotor attention to the letter was larger for the efficient group than for the inefficient or control groups. This means that, although the inefficient group made more foveations overall than the efficient group, they spent fewer foveations on the letter relative to other objects than did the efficient group. The finding that the proportion of foveations a subject directed to the letter was correlated positively with game score suggests that the effect of oculomotor efficiency not only is related to a general reduction in the number of foveations, but also to the appropriateness of objects foveated. In the no-mine periods of the game, an efficient use of eye movements would dictate minimizing foveations to the display indicating fortress hits, since this display would need to be viewed only when the number of hits approached the required number. By the end of the experiment, the efficient group viewed the display once on approximately one out of every five no-mine periods, while the inefficient and control groups viewed the display twice as often. Since the efficient group had significantly more hits on the fortress than the inefficient or control groups on the last block of the experiment, such a finding supports the notion that efficiency of eye movements is related to performance. The eye movement patterns of the inefficient group and the control
240
K. L. Shapiro,
J. E. Raymond
/ Traming
oculomotor
strategres
group generally were very similar, at least by the last session of the experiment. It is likely that the inefficient drills either were ineffective at modifying oculomotor behaviour or that the acquired oculomotor habits were not transferred to the complex task of the whole game. Evidence in support of the latter possibility can be seen in both panels of fig. 3. While efficient and inefficient groups began with approximately the same number of foveations per game period, after approximately 100 min in the experiment (after several sessions of drill practice), the inefficient eye movement group exhibited a sharp increase in the total number of foveations while the efficient group did not. This was particularly so for the mine periods of the game where objects appear in the periphery and therefore drill training might be expected to have greater impact than in the no-mine periods. On subsequent eye movement recording sessions, both groups show a decline in the number of foveations, suggesting that the inefficient group abandoned the eye movement strategy imposed by the drills in the latter stages of the experiment. A similar pattern of foveations can be seen in fig. 4 which illustrates the change in foveations to the letter region during mine trials with time in the experiment. Here again, the inefficient group shows an increase in the number of foveations after approximately 100 min the experiment. Relative to the initial level, there was a two-fold increase in the number of eye movements for the inefficient and control groups suggesting that the inefficient drill training did impact oculomotor behaviour early in training. One of the underlying ideas of the present investigation is that subjects develop when confronted
oculomotor patterns which are applied repeatedly with a similar situation. Evidence that such oculomo-
tor habits were formed is found in the analysis of eye movements during subjects’ attempts to hit either the mine or the fortress. We found that most subjects (about 70%) adopted specific styles of viewing either: (1) the ship, or (2) the object of the attack immediately before firing. The remaining 30% did not display a specific habit but alternated their gaze between ship and object just before firing. In general, the subjects that did not develop a fixed habit had the lowest scores in the subject sample suggesting that habit formation is helpful in performance. For many years it has been demonstrated that the training of any complex perceptual motor task benefits significantly from a part-task training regimen. Part-task training involves taking a complex task and
K.L. Shapiro, J. E. Raymond
separately
training
each
of
/ Training oculomotor strategies
the
important
component
241
parts
before
reconstructing the part tasks into the whole task. Typically, part-task training has included components such as cognitive and motor performance features. While the visual features of a task usually are considered to be important, part tasks typically are not designed with the oculomotor requirements of the whole task in mind. The results of our present research suggest that there are two active components of the visual perceptual system which can be brought under voluntary control by incorporating each into an oculomotor part-task training regimen: (1) the oculomotor system, which acts to place both moving and stationary objects on the fovea, and (2) the visual attentional system which acts to maximize the use of fovea1 and peripheral visual sub-systems. The visual attentional system is important for inputing visual information from each of these two components of the visual system in selective situations where simultaneous input is necessary. For example, fovea1 input is required to manipulate an object requiring fine motor control, while the peripheral system is best suited for detecting the appearance of peripheral objects which require only that their presence be detected. The results of our investigation suggest that oculomotor task requirements in addition to the more traditionally considered motor task requirements should be analyzed when designing part-task training regimens. Such considerations may enhance the efficacy of part-task training procedures by pointing out a broader scope of perceptual motor behaviours contributing to complex perceptual motor task training.
References Bahill, T. and T. LaRitz,
1984.
Why can’t batters
keep their eyes on the ball? American
Scientist
72, 249-253. Birch, H. and A. Lefford,
1967. Intersensory
development.
Monographs
of the society for research
in child development. Clement,
W.,
instrument
H.
Jex
and
D.
Graham,
landings of a jet transport.
1968. IEEE
A
manual
Transactions
control-display
theory
on Man-Machine
applied
Systems
to
MMS-9.
93-110. Eriksen,
C.W. and J.E. Hoffman,
from visual displays. Festinger,
1972. Temporal
Perception
L., 1971. ‘Eye movements
(eds.), The control
and perception’.
of eye movements.
Fitts, P.M., R.E. Jones and J.E. Milton, landing approaches.
Aeronautical
and spatial characteristics
and Psychophysics New York: 1950.
In: P. Bach-y-Rita, Academic
Eye movements
Engineering
of selective encoding
12, 201-204. C. Collins and J.E. Hyde
Press. pp. 259-341. of aircraft
Review 9, l-5.
pilots during instrument
242
K. L. Shapiro, J.E. Ruymond
Howard,
I.P.,
1971.
Mechanisms Jeannerod,
‘The
adaptability
of motor
M., P. Gerin
J.W.
and J. Pernier,
1968.
Vision
1978. ‘Studies
and R.A. Monty
functions.
Hillsdale,
E. and J.W.
Pensacola, Ludvigh,
FL,
NJ:
Naval
1954.
School
Miller,
1. Introduction.
In:
K.J.
Connolly
(ed.),
Press. pp. 337- 352.
et fixation
du regard
dam
l’explora-
8, 81-97.
of eye movements
and visual search in radiology’.
(eds.), Eye movements
In:
and the higher psychological
Erlbaum.
Miller,
E. and J.W.
objects.
Fisher
srrategies
system’.
Academic
Deplacements
Research
H. and C.F. Nadine, D.F.
visual-motor
London:
Senders,
Ludvigh,
Mane.
the
skill development.
tion libre d’une scene visuelle. Kundel,
of
/ Traimng oculomotor
Some effects fo Aviation
1958.
on dynamic
visual acuity.
Study of visual acuity during the ocular
Journal
A. and E. Donchin.
of training
of the Optical
1989.
The
NSAM-567.
Medicine.
Space
Society
Fortress
of America
game.
Acta
pursuit
of moving
test
48, 7999802.
Psychologica
71. 17-22
(this
volume). Matin,
E.,
1974.
Saccadic
suppression:
A review
and
an analysis.
Psychological
Bulletin
81.
8999917. Moray,
N., 1984. ‘Attention
man and D.R. Norton,
to dynamic
visual displays
Davies
(eds.),
Varieties
D. and L. Stark,
1971.
Scanpaths
of attention.
in manmachine New York:
in eye movements
systems’.
Academic
during
pattern
In: R. Parasura-
Press. pp. 485-512. perception.
Science
171,308311. Posner,
M.I.,
1980.
Orienting
of attention.
Quarterly
Journal
of Experimental
Psychology
32.
Ergonomics
16,
3-25. Schoonard.
J.W.,
J.D.
Gould
and L.A. Miller,
1973.
Studies
of visual inspection.
365-379. Senders,
J.W.,
1983.
Visual
Senders,
J.W.,
J.E.
Elkind,
visual sampling Williams,
H., 1974.
R. Martens Kinetics Yarbus,
A.L.,
behavior
scanning M.C.
processes.
Grignetti
of human
‘Perceptual-motor
and M. Wade (eds.),
Tilburg:
and R.A.
observers.
of Tilburg 1964.
as a function
of motor behavior
of information
Eye movements
and vision.
of the
New York:
processing’.
and sport. Urbana,
Publishers. 1967.
Press.
An investigation
NASA-CR-434.
development Psychology
University Smallwood.
Plenum
Press.
In:
IL: Human