You are on page 1of 28

APPLIED COGNITIVE PSYCHOLOGY, VOL.

13, 127 (1999)

A Multidimensional Approach to Skilled Perception


and Performance in Sport
WERNER F. HELSEN1* and JANET L. STARKES2
1

Katholieke Universiteit Leuven, Belgium


2
McMaster University, Canada

SUMMARY
Three experiments examined the relative importance of attributes determined largely by the
eciency of the visual/central nervous system versus cognitive domain-specic skills, in the
determination of expertise in soccer. In Experiment 1, expert and intermediate soccer players
were assessed on various non-specic abilities including: processing (simple reaction time,
peripheral reaction time, visual correction time), optometric (static, dynamic and mesopic
acuity), and perimetric parameters (horizontal and vertical peripheral range). In Experiment 2,
domain-specic variables were assessed including complex decision speed and accuracy,
number of visual xations, xation duration, and xation location in solving game problems.
Stimuli were initially presented by slides (Experiment 2) and later by 16 mm lm (Experiment 3). Eye movements were recorded and analysed. A stepwise discriminant analysis of
both non-specic abilities and soccer-specic skills revealed an average squared canonical
correlation 0.84, with the signicant step variables all being domain-specic skills.
Copyright # 1999 John Wiley & Sons, Ltd.
Appl. Cognit. Psychol. 13: 127 (1999)

The study of experts is rapidly expanding in areas such as cognitive psychology and
cognitive science (Ericsson and Lehmann, 1996). Likewise the study of `motor'
experts has ourished recently (Abernethy, 1991; Ericsson, 1996; Starkes and Allard,
1993). One longstanding debate in the expertise research has been the question of
whether experts are born or made. This nature/nurture issue has been examined
extensively elsewhere (Ericsson, 1996; Ericsson and Lehmann, 1996; Howe, Davidson
and Sloboda, in press). In motor research the question remains whether abilities are a
precursor of skill, or whether skill is the result of extended practice in a specic
domain.
If one examines the history of research on motor experts in sport, the period from
the 1950s to early 1980s is characterized as period of search for innate individual
dierences in either optometric or processing abilities. The basic assumption was that
the skilled athlete must have superior levels of vision, or perceptual-motor speed to
excel in tasks that supposedly are compilations of these component skills. Thus, it was
common to study the relationship between non-task-specic visual components and
*Correspondence to : Dr. Werner Helsen. Katholieke Universiteit Leuven, Motor Learning Laboratory,
Tervuursevest 101, 3001 Leuven, Belgium.

CCC 08884080/99/01000127 $17.50


Copyright # 1999 John Wiley & Sons, Ltd.

Accepted 26 November 1997

W. F. Helsen and J. L. Starkes

one's absolute performance level in sport. Along the way various optometric
parameters were invoked static visual acuity, dynamic acuity, stereoacuity, and
peripheral range as most predictive of skill. Among processing variables, such
abilities as simple reaction time, peripheral reaction time and nerve conduction
velocity were proposed as potential sources of superiority. An in-depth analysis of the
component ability approach is covered in the original work of Fleishman (1966) and
the publications of Keele and Hawkins (1982) and Allard and Starkes (1991).
Although the relationship between these ability components or what has been
elsewhere termed, `hardware' and sport skill is logically and intuitively appealing, this
relationship has not accounted for a signicant amount of variance in skilled
performance (Starkes, 1987; Wrisberg, 1993).
A brief summary of the various ndings on sport experts to date suggests that:
(1) Expert athletes retain, recall, and recognize signicantly more information about
structured game situations than less experienced participants, when information
is domain-specic and presented briey. This is true in sports such as basketball
(Allard, Graham and Paarsalu, 1980), eld hockey (Starkes, 1987), volleyball
(Allard and Starkes, 1980; Ripoll, 1988), and soccer (Helsen and Pauwels, 1993a).
(2) Motor expertise is reected in greater amounts of declarative knowledge about
one's domain (French and Thomas, 1987; McPherson, 1993), and this knowledge
has been shown to be a constituent skill rather than a byproduct of time spent in
the domain (Allard et al., 1993; Williams and Davids, 1995).
(3) Protocol analyses reveal that sport expertise is reected in greater amounts of
declarative, procedural, and strategic knowledge of one's domain (French and
Thomas, 1987; McPherson and Thomas, 1989; McPherson, 1993; Rutt Leas and
Chi, 1993). There is also some indication that young experts, who are rich in
declarative knowledge and perform well on cued recall of domain related
information, also benet from more accurate estimates of their own recall abilities
provided by metacognition (Glaser, 1996; Schneider and Bjorklund, 1992).
(4) Superior performance in motor experts doesn't seem to be linked to dierences in
underlying visual or sensory abilities (Starkes et al., 1995). There is still controversy, however, whether there are inherent dierences in experts related to
complex decision making, force control, or attentional processes (Keele, Ivry and
Pokarny, 1987).
(5) Motor experts usually deal with domain-related information more quickly and
eciently than less skilled individuals. This is evident from their use of advance
visual cues for decision making (Abernethy and Russell, 1984; Goulet, Bard and
Fleury, 1989; Starkes et al., 1995; Wright, Pleasants and Gomez-Meza, 1990). It
is also apparent from research on the measurement of eye movements and the
number of xations, xation location, duration, and scanpaths of sports experts
(Abernethy and Russell, 1987; Bard and Fleury, 1976; Goulet et al., 1989; Helsen
and Pauwels, 1993a,b).
(6) Finally, Abernethy and Russell's studies (e.g. 1987) on visual search strategies in
a racquet sport (using eye movement analyses in a xed head/dynamic lm task)
have shown that it is not so much the manner in which the information is overtly
searched but rather the use the performers can subsequently make of this available information that produces the dierence between the perceptual performance of the expert and that of the novice.
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

Expertise and visual information processing

While these ndings suggest a great deal about superior performance in sport, there
is one drawback common to most of the research. Each of the studies has assessed
performance in one sport, and usually on a limited number of dependent variables.
To date only Starkes (1987) and Abernethy, Neal, and Koning (1994) have attempted
to assess performance on a number of visual, chronometric, signal detection, recall,
and complex decision tasks. In those studies the goal was to evaluate the multivariate
contribution of various perceptual and cognitive components in predicting expert
performance.
Starkes (1987) assessed three task-independent component abilities (visual reaction
time, dynamic visual acuity, and coincident anticipation time) as well as a number of
eld hockey-related skills (shot prediction before and after view of ball impact, signal
detection of a ball in game slides, decision of optimal oensive moves in briey
exposed slides, and recall of player positions seen in slides). Her results indicated a
multiple R 0.83 (r2 0.69) with two signicant step variables, both being `software'
or domain-specic variables. The primary variable was recall of game-structured
information. The second variable was shot prediction accuracy following view of
ball impact. No other variables contributed signicantly in the prediction of group
membership for expert, intermediate or novice players. The reported ndings clearly
provide support for the contention that skill dierences can be attributed to the
perceptual/cognitive aspects of a task.
However, Starkes did not assess all the possible visual abilities potentially
suggested as important in sport, and game information was provided largely by static
stimuli (slides) that required only verbal responses. Nevertheless the study illustrates
the need for multi-task analyses of expertise within subjects. Wrisberg (1993) has
emphatically suggested that no single factor or ability will account for performance
dierences in sport, and that research needs to examine the particular grouping of
factors that contributes most to skilful performance in particular domains. Predictions of expertise need to be interactionist and multidimensional. Eventually by
determining those components that pose consistent diculty for a particular athlete,
specic intervention or training programmes could then be implemented.
This series of studies of soccer players assess a large range of processing, optometric, and perimetric parameters, and implements soccer-specic perceptual,
decision and motor tasks using both static and dynamic stimuli. Table 1 provides a
synopsis of each of the factors assessed over the three experiments.
Many of the tasks presented in these experiments have been reported individually
elsewhere (Helsen and Pauwels, 1988, 1990, 1993a,b; Helsen, Pauwels and Van
Outryve d'Ydewalle, 1986) but this is the rst attempt to take two groups of soccer
players (one more and one less skilled) and examine their performance across the full
range of tasks designed to tap rst a variety of non-specic abilities related to visual/
central nervous system function and then increasingly soccer-specic skills. The
strength of the paper is that like the earlier work of Starkes (1987) and Abernethy et al.
(1994) multivariate analyses permit an examination of the relative contribution of
each underlying factor in predicting soccer skill.
Experiment 1 is a test of expert versus intermediate soccer players for dierences in
basic visual and reactive capacities. Experiment 2 tests performance dierences using
more real-world stimuli (soccer slides). Experiment 3 is an attempt to capture the
dynamic information and impose the time constraints aorded in a real soccer game
by assessing dierences using 16 mm lm stimuli. As one moves from Experiment 2 to
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

W. F. Helsen and J. L. Starkes

Table 1.

Variables used in the prediction of soccer expertise

NON-SPECIFIC ABILITIES
Processing parameters
Central reaction time
Peripheral reaction time to stimuli at 908
Peripheral reaction time to stimuli at 1808
Visual correction time
Static display
Dynamic display
Optometric parameters
Static acuity
Near
Intermediate
Far
Mesopic
Dynamic acuity
Perimetric parameters
Horizontal dimension
Vertical dimension

SOCCER-SPECIFIC SKILLS
Perceptual parameters
Number of visual xations
Fixation duration
Fixation location
Decision parameters
Response accuracy
Motor parameters
Initiation time
Movement time
Ball-contact time
Ball ight time
Total response time

|{z}

for both
Static display Dynamic display
(slides)
(16 mm lm)

Experiment 3, if dierences are heightened based on delity of the task to real soccer
the assumption is that domain specicity is most important in predicting expertise.
EXPERIMENT 1: LABORATORY TESTS OF NON-SPECIFIC ABILITIES
The question examined in Experiment 1 is whether expert and intermediate soccer
players dier on performance of processing, optometric, and visual perimetric
components.
Method
Subjects
The same 28 male subjects voluntarily participated in all three experiments. The
expert group (mean age 26.3 years) consisted of 14 semi-professional players
who averaged ten years of prior active competition. The intermediate group (mean
age 22.5 years) consisted of 14 undergraduate kinesiology students. They had
experience with physical activity and possessed some skill in soccer to be able to
perform Experiment 3 (as veried in a preliminary interview and pre-experimental
session). The intermediate group, however, had no experience in competitive soccer
beyond high school. In pre-experimental interviews it was determined that none of the
kinesiology group had extensive experience or skill in any interceptive or ball sport.
Informed consent was obtained from all subjects prior to participation.
Apparatus and tasks
Processing parameters
. Central and peripheral reaction time. The subject was seated in front of a large
custom-made hemisphere which surrounded the entire upper body. The head was
placed in a frontal head support so vision was focused in the middle of the hemisphere. The hemisphere had embedded in it 24 light sources (12 red and 12 green)
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

Expertise and visual information processing

arranged in a semi-circular fashion. To measure central reaction time a green light


source just in front of the subject was lit up from 1.1 to 1.8 s after a red foreperiod
signal (also located in front of the subject). At the onset of the stimulus, the subject
pressed a telegraph key as quickly as possible with the index nger of the preferred
hand. Timing was stopped as soon as the light was turned o. After ve practice
trials, the subject was given two series of ten trials, each separated by a 2-minute
rest. Using the same test set-up, subjects then had to react to stimuli presented by
peripheral light sources at 908 and 1808, both left and right. After ve trials, the
subject was given two series of 20 trials to evaluate peripheral reaction time. Both
series were separated by a 5-minute rest and divided equally between the dierent
light sources. The sequence of stimuli was randomized.
. Visual correction time. First presented by Abernethy and Russell (1987), visual
correction time (VCT) reects the minimal time required to initiate the rst saccadic
eye movement in response to a stimulus. In this study, VCT was the time between
the onset of a lm sequence and the actual initiation of the rst eye movement. VCT
in our experiment provides information regarding the speed with which visual
search is initiated in response to slide (Experiment 2) and 16 mm lm projection
(Experiment 3).
Optometric parameters
. Static visual acuity: near, intermediate, far, mesopic. Static visual acuity or the ability
to detect detail of an object was tested at dierent distances. The subject placed the
head against the frontal support of a cylinder. Charts ranging from 12 to 2 minutes
of arc were presented individually. Each chart had two lines of optotype which
included 4 letters and 4 digits. The examination started with acuity level 10/10 and
if the subject had at least three answers right (for letters and digits), it progressed to
12/10. If the subject was not correct, stimuli were increased in size until the subject
correctly identied three items. Near vision was checked at 33 cm (3 diopters of
accommodation) on an axis of 518 below horizontal. Intermediate vision was
examined at 66 cm on an axis of 338 below horizontal; while far vision was checked
on an axis of 158 below horizontal. To test mesopic vision, the acuity scale appeared
in low luminance, i.e. 2 cd/m2.
. Dynamic visual acuity. Essentially tests of dynamic visual acuity (DVA) evaluate the
subject's ability to predict the relative velocity of a moving object and adjust the
ocular system to `catch' and hold the image of the object in the vicinity of the fovea
long enough to perceive object detail. Morris (1977) provides an excellent summary
of DVA tests available, and the likely role DVA plays in ball sports. Our equipment
is comparable with the apparatus used by Sanderson and Whiting (1978), and
Starkes (1987). A chevron moved across from left to right in front of the subject, and
the subject had to report whether the point of the chevron pointed to the right or to
the left. On each trial the image moved past a viewing slit in the cylinder once. Three
dierent speeds were combined with three dierent acuities (0.3, 0.2, 0.1). At each
speed, subjects were forewarned of the approach of the stimulus. Stimulus speeds
(10, 19, and 24 cm/s) were chosen to best reect the relative speeds routinely
required to visually track a soccer pass or shot. Subjects performed nine randomized
chevron position trials within each of the three speed conditions. Order of presentation of speeds was randomized across subjects.
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

W. F. Helsen and J. L. Starkes

Perimetric parameters
. Dierential light threshold. A spherical projection perimeter (Goldmann Perimeter
940) was used with a system permitting direct registration of the target position.
Illumination of the sphere (radius) was constant and uniform, so that adaptation
of the whole eye was identical. To determine the dierential threshold of light
sensitivity in central and peripheral parts of the visual eld, a target was projected
onto the inside of the sphere in both vertical and horizontal directions. The object
size was 1 mm2 and its relative intensity level 0.315. The luminosity of the sphere
(31.5 asb; 1 asb 0.1 millilamberg mL) was adjusted by a shutter, and the brightness of the target (1000 asb) by a light-meter and a rheostat. Both were illuminated
simultaneously by a single bulb.
The target detection task required a rapid button press with a stylus held in the
preferred hand. A built-in electromagnet activated a marker on the recording device.
This procedure enabled the examiner to keep a constant target speed and provided an
automatic registration of target position. The xation of the subject's eye could be
constantly checked through a reticulated telescope. Each subject was required to sit in
front of the perimeter and keep one eye focused on a xation point throughout the
trial. In the vertical direction, two targets were given at 758, 908, 1058, 2558, 2708, and
2858. In the horizontal direction, two targets were given at 08, 158, 1658, 1808, 1958,
and 3458. The sequence of all targets was randomized. A 5-minute rest was given
between the testing of each eye.
Procedure
Standardized instructions were read to each subject. Order of presentation of the
dierent tests was randomized across subjects.
Design
To assess non-specic abilities, the following dependent variables were assessed: for
the processing parameters, reaction time (central, peripheral 908, peripheral 1808),
and correction time (static and dynamic display); for the optometric parameters,
static acuity (near, intermediate, far, mesopic vision), and dynamic acuity; and for
the perimetric variables horizontal and vertical peripheral visual range. Mean
performances were calculated separately for each task to compare groups. A two-way
analysis of variance (ANOVA) was carried out on these means. G-G Epsilon values
were also assessed and since they were all greater than 0.75 no correction was
employed (Schutz and Gessaroli, 1987). As such, regular probability values are
reported. The sources of any signicant main eect were identied through Tukey
HSD post-hoc procedures, where appropriate. For signicant interactions, simple
main eects were analysed using F-tests. All statistical tests were completed with
alpha set at 0.01.
While much of the literature reviewed previously would predict no dierences
between experts and intermediates, the test hypothesis predicted that experts would
perform better than intermediate-level players on reacting to a visual stimulus,
initiating the search sequence in response to a static and dynamic display, their
optometric abilities, and perimetric capacities.
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

Expertise and visual information processing

Table 2. Mean values (in ms) and standard errors (in parentheses) of central and peripheral
reaction times and visual correction time for both groups
Reaction time (ms)

Visual correction time (ms)

Group

Central

Peripheral 908

Peripheral 1808

Slides

Film display

Expert

216
(5.16)

244
(4.67)

278
(6.36)

226
(5.83)

225
(6.21)

Intermediate

221
(4.32)

253
(4.88)

288
(5.59)

237
(5.15)

228
(5.71)

Results and discussion


Processing parameters
Central and peripheral reaction times.
The available data in Table 2 suggest that reaction times were within the range
reported elsewhere in the literature (Starkes and Deakin, 1984). A 2  3 (Expertise 
Stimulus light position) ANOVA with repeated measures on the last factor only
revealed a signicant main eect for Condition (08, 908, and 1808), (F(2, 52) 188.14,
p 5 0.0001). Groups were not signicantly dierent. Post-hoc procedures showed
an overall signicant eect of peripheral presentation of the stimulus on reaction
time, which declined progressively from 1808 (M 283 ms+ 4.96) over 908
(M 249 ms+ 3.43) to central presentation (M 219 ms+ 3.34). Even in these
simple tasks the reaction to a visual stimulus took longer when the stimulus was oered
peripherally.
The argument is an old one, that reaction time is inherently linked to elite
performance in sport. The lack of correlation between reaction time and movement
time, however, suggests that a fast reaction time is probably not an underlying
component of good performance (Starkes and Deakin, 1984). Second, those studies
that have attempted to correlate simple reaction time and sport skill have shown
inconclusive results, some authors showing a positive relationship, while others not
(McLeod, 1987). Our results, given in Table 2, conrm this second line of evidence,
showing little empirical evidence for a conclusive link between reaction time and
soccer expertise.
Visual correction time.
The magnitudes of the VCTs observed (200250 ms) for both the expert and intermediate group suggest that the delay in commencing the initial saccadic movement
represents a typical simple reaction time delay. This concurs with the high eye movement latencyreaction time correlations reported for other selective eyehead
coordination tasks (Abernethy and Russell, 1987; Gangemi et al., 1993). A 2  2
(Expertise  Display) ANOVA with repeated measures on the last factor revealed no
signicant dierences. Although expert subjects displayed more rapid VCTs than
intermediates, neither dierences in subject prociency nor the slide versus lm
condition signicantly inuenced the time taken to make the rst saccadic response
to the display. Viewed in this light, the observation of systematically, but not signicantly, faster VCTs for the expert performers may be as expected (Yandell and
Spirduso, 1981). The relatively low variability observed in the VCTs also supports the
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

W. F. Helsen and J. L. Starkes

notion that this parameter is a reection of an inherent constraint within the visual
system.
Optometric parameters
Static visual acuity: near, intermediate, far, mesopic
A 2  4 (Group  Vision condition) ANOVA with repeated measures on the last
factor only revealed a signicant main eect for the near, middle, far and mesopic
vision condition (F(3, 78) 21.84, p 5 0.0001). The dierence between groups was
not signicant. This conrms the ndings of earlier studies (Morris, 1977; Sanderson,
1981). Post-hoc procedures showed a signicant dierence between mesopic vision
(M 9.29+ 0.46) and near (M 11.57+ 0.29), middle (M 11.57+ 0.21) and far
vision (M 11.00+ 0.39).
Dynamic visual acuity
Analysis of DVA showed that the average subject made less than 5% error. Like
earlier studies (Morris and Kreighbaum, 1977; Starkes, 1987), a ceiling eect occurred
for this task since subjects were able to perform even the fastest speed with minimal
error. There was no advantage for the expert group. In fact, at 24 cm/s the intermediate group performed better (Mean intermediate (Mi) 95.24%+ 1.26) than the
experts (Mean expert (Me) 85.71%+ 2.53).
Perimetric parameters
Dierential light threshold
The results in horizontal (Me 167.188+ 1.73; Mi 156.878+ 2.18) and vertical
dimensions (Me 102.378+ 1.74; Mi 96.308 + 1.80) suggest that data are within
the range previously reported in the literature (Williams and Thirer, 1975). Dierences between the groups in the horizontal dimension were signicant (F(1, 26)
13.72, p 5 0.001); while dierences in the vertical dimension were not. To date,
discrepant ndings have occurred in athlete, non-athlete comparisons of peripheral
vision. Some authors have reported signicant dierences in the horizontal or the
vertical dimension, while others have not. Our results are in line with the more recent
ndings (Doil and Binding, 1986; Mizusawa, Sweeting and Knouse, 1983), showing
signicant dierences in the horizontal but not vertical dimension.
Conclusions
From this study of the relationship between dierent non-specic abilities and soccer
skill it is apparent that the notion of the skilled athlete as a well-constructed conglomerate of visual information processing, optometric, and perimetric components
is not a productive stance. The expert players tested possessed neither superior central
and peripheral reaction times nor superior visual correction times. Skilled soccer
players did not dier from less experienced ones in their static visual acuity or their
ability to visually track a moving object (at least up to a rate of 24 cm/s). Finally,
players had only average vertical peripheral visual range. In earlier studies a lack of
dierence may have resulted from lack of precision in previous technology. In this
study, however, with the most up-to-date technology, dierences were still not
apparent. Nevertheless, the argument that these parameters are inherently linked to
elite performance is a long-held perception of coaches, and a dicult one to change.
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

Expertise and visual information processing

The next logical question was to what extent expert performance may be guided
by the processing of sport-specic perceptual information. Experiment 2 examines
whether skill dierences emerge in a soccer decision task presented by static slide
stimuli.
EXPERIMENT 2: EYE MOVEMENT ANALYSES OF SOCCER SLIDES
A well-documented observation from the original research of Chase and Simon
(1973) on chess experts, which holds true in certain team sports (for a review see
Helsen and Pauwels, 1993b), is that experts outperform novices in the recall of briey
presented domain-specic stimulus information (e.g. slides of basketball attacks in
progress). However, this is not true for displays lacking domain-specic game
structure (e.g. slides of basketball players positioned randomly, such as rebounds
turnovers or a scramble for a loose ball).
As a number of researchers have argued, this selective superiority of experts in the
recall of structured display information is not so much a function of dierences in
pattern recognition capacity as it is a reection of expertintermediate dierences in
knowledge (e.g. Holding and Reynolds, 1982). In sports too, experts have been shown
to have greater knowledge than non-experts in terms of knowing both specic facts
about their sport (declarative) and how to select correct courses of action (procedural)
in their sport (French and Thomas, 1987; McPherson, 1993). Indeed, across the
duration of a playing season, improvements in performance of young players may be
more attributable to knowledge development than to changes in skill (French and
Thomas, 1987).
Another important attribute of skilled chess players in these tasks seems to be the
ability to view large amounts of structured information, reduce it through various
knowledge compilation and composition processes (such as chunking), and briey
retain that product in working memory. The perception of complex displays in terms
of chunks of meaningful associations rather than a compilation of individual items
(i.e. individual player positions) both speeds up the processing rate and facilitates a
deeper level of processing; an action which, in turn, facilitates selective superiority of
experts in recognizing and recalling structured display information.
The object of the second more domain-specic study was to analyse this process of
information reduction by means of eye movement registration. This technique was
originally applied in sport by Bard and Fleury (1976), who showed that selection of
information in problem-solving situations in basketball was characterized by a
coherent perceptual strategy. Soccer, like many team sports, involves strategies to
score goals, or gain space from opponents in an optimal and systematic way. Other
sports (such as basketball, football, volleyball and eld hockey) have similar strategy
requirements. In soccer, a team has 11 players, each of them with very specic
defensive and oensive roles and positions. The large size of the playing eld dictates
that any structured attack is apparent, at least to the trained eye! Moreover, it is an
open sport, performed in an environment which is rapidly changing and in which
both perceptual uncertainty and time-constrained decision making are critical
features. As Allard and Burnett (1985) have suggested, open-skill athletes probably
oer the best prospect for assessing the cognitive requirements of a sport, given the
need for strategy and structure in play.
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

10

W. F. Helsen and J. L. Starkes

From previous results using the recall paradigm, it was hypothesized that soccer
expertise would be characterized by an increase in selectivity, and thus shorter
processing and response times.
Method
Subjects
All subjects in this experiment had also participated in Experiment 1. These were 14
male semi-professional soccer players and 14 male kinesiology undergraduates.
Apparatus and task
Each subject was presented 30 slides, showing a typical oensive soccer situation
(80* 50 cm, 708 of visual angle), from the point of view of the player with the ball. The
process to establish 30 criterion slides was extensive. Fifteen coaches at the semiprofessional level examined a large pool of slides and determined which represented
situations where there was only one optimal oensive move (to shoot, pass, or dribble
the ball). Only slides for which there was 90% agreement across coaches were kept as
a test stimuli.
Within the task a constant preparatory signal was given before presentation of the
slide. At slide onset the player verbalized his choice of the optimal oensive move. He
was instructed to do so as quickly and accurately as possible. The possible answers
were: shoot at the goal, dribble around the goalkeeper or an opponent, or pass to one
of four team mates. Timing was stopped as soon as the subject responded. A custommade timing apparatus started a clock with slide onset and recorded vocal reaction
time when the subject verbalized his response. Reaction time was recorded in
milliseconds.
As illustrated in Helsen and Pauwels (1990), eye movements were recorded by a
NAC-V Eye Movement Recorder. In this system, a light ray is reected from the
anterior surface of the cornea. The scene-camera records the eld of view of the
subject, while the eye-cameras record the movement of the eye. Fixation location is
superimposed on the scene the subject views. Both are recorded on videotape and
simultaneously displayed onto a high-resolution monitor placed out of the subject's
eld of view. As in previous work (Helsen and Pauwels, 1990), the accuracy of the eyemovement recorder was less than 28 of visual angle, and data were sampled at 50 Hz.1
Procedure
The order of presentation of the slides was randomly assigned. The experimental setup was explained individually to each subject by means of standardized instructions.
The eye-movement recording apparatus was tted onto the subject's head and
stabilized. It was then calibrated for both position and linearity to ensure that the
xation mark (a light spot reected from the subject's dominant eye) corresponded
precisely to the subject's visual orientation to dierent sectors of the projection
screen. The subject was further instructed to return visual focus to the screen centre
immediately after completing his response. Whenever this visual orientation was not
apparent from the monitored eye movements, the test was stopped, and the apparatus
1

While 50 Hz is a low sampling frequency, elsewhere it has been shown that for these types of studies the
data acquired is sucient and similar to that acquired by higher-frequency (120 Hz) systems (Helsen et al.,
in press).

Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

Expertise and visual information processing

11

recalibrated. Five slides were used for familiarization, followed by 30 game slides,
equally divided in two series of fteen. Both series were separated by a 2-minute rest.
Design
In addition to level of expertise, the type of problem-solving task was taken into
account as a second independent variable, incorporating three types of solution:
shoot at the goal (N 10); dribble around the goalkeeper or an opponent (N 8);
and pass to a specic and free team mate (N 12). Five dependent variables were
assessed. Voice reaction time and response accuracy provided response measures. For
evaluation of the ongoing cognitive processes, the visual search patterns were
analysed frame by frame. Ocular responses were described in terms of both location
and duration. To derive location data the display was divided into discrete cues, such
as the goal, the player in possession of the ball, or the free team mate. Fixations which
were either to an unnamed region of the display or to a location which could not be
clearly identied were designated `rest'. An `eye xation' was dened as a period that
lasted at least 140 ms, and in which the eye movements were limited to an area that
corresponded in angular size with the fovea. These denitions correspond to previous
literature (Bard and Fleury, 1981; Helsen and Pauwels, 1990). Given a video sampling
rate of 50 frames/s, these coded input data were then used to compute the following
series of dependent measures: number of xations, xation duration (average
duration of all xations occurring during the slide's appearance), xation location
(percentage of the trial time allocated to each xation location outlined previously). A
two-way ANOVA was carried out for each dependent measure, followed by the Tukey
HSD post-hoc tests outlined previously.
The question examined in Experiment 2 was whether expertintermediate dierences emerge in a soccer decision task. More specically, the study was designed
to determine (1) if experts were faster and more adequate in nding tactical solutions;
(2) if experts diered from less experienced players in their selectivity and processing
speed; and (3) if subjects used dierent visual cues according to the type of solution.
While the use of slide technology may seem dated, recall that the purpose of
presenting static stimuli was to later test the improvements in performance aorded
by more realistic stimuli in Experiment 3.
Results and discussion
Response speed and accuracy
The analysis of response time data (Table 3) showed that the two main eects, Group
(F(1, 26) 8.93, p 5 0.0061) and Type of problem-solving task (F(2, 52) 21.12,
p 5 0.0001), were signicant. Post-hoc tests revealed that the shotting situations were
solved signicantly faster than dribbling (p 5 0.01) and passing situations (p 5 0.01).
Analysis of response accuracy showed that the two main eects were signicant, as well
as the interaction eect (F(2, 52) 6.91, p 5 0.002). Shooting situations were solved
more accurately than either dribbling (p 5 0.01) or passing situations (p 5 0.01).
Post-hoc tests further revealed that experts performed better than intermediates for
dribbling ( p 5 0.01) and passing situations (p 5 0.01) but not for shooting situations.
When the response times for this task were compared with earlier central and
peripheral reaction times, the complexity of the shoot/dribble/pass decision was
veried.
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

12

W. F. Helsen and J. L. Starkes

Table 3. Mean values and standard errors (in parentheses) of response time (ms) and
response accuracy (%) for both groups in solving shooting, dribbling and passing situations,
presented by means of slides
Response time (ms)

Response accuracy (%)

Group

Shooting

Dribbling

Passing

Shooting

Dribbling

Passing

Expert

1081
(59.27)

1332
(80.74)

1560
(98.04)

97
(1.63)

82
(4.29)

82
(3.08)

1502
(133.22)

2070
(269.20)

2064
(147.89)

95
(1.74)

57
(5.82)

59
(4.27)

Intermediate

These results conrm previous ndings in team sports, that response speed and
accuracy are dependent upon competence, with experts showing faster response times
and higher response activity accuracy than intermediates in solving tactical game
problems, presented statically by means of slides (Bard and Fleury, 1976; Pauwels and
Helsen, 1986). According to the results found in a similar basketball task (Bard and
Fleury, 1976, 1981), the type of problem-solving task signicantly inuences response
time and response accuracy; shooting situations are solved faster and more accurately
than dribbling and passing situations respectively.
Finally, by contrasting data obtained from Experiment 1 (of response speeds for
central and peripheral RT) with responses to the soccer slides in Experiment 2 we were
able to examine whether the robust expert/intermediate interaction usually found for
game structured stimuli in recall paradigms also holds for these tasks. A two-way
ANOVA was conducted for Group (experts/intermediates) by Type of response
(central RT, slide decision time). This same analysis was also conducted for peripheral
RT, as a test of the same issue. Since results for the two contrasts were very similar,
only the rst ANOVA using central RT and slide decision time will be presented.
The results showed an interaction very similar to that found in experiments in
which experts and non-experts had to recall game structures and non-structured
situations. That is, there was a signicant interaction (F(1, 26) 10.14, p 5 0.0037)
such that both skill groups were similar for the non-specic central RT (Me
216 ms+ 5.16; Mi 221 ms+ 4.32). A dierence emerged for the soccer slides
(Me 1324 ms+ 54.87; Mi 1878 ms+ 116.51). Even though this comparison is
made between a manual and a vocal reaction time, and therefore may be due simply to
the nature of the response required, it is interesting that when structure is imposed by
the nature and specicity of the decision task in sport the expert/intermediate
interaction begins to emerge. Again this is similar to the results typically shown within
the recall paradigm for game-structured versus non-structured information.
Number of xations and xation duration
The analysis of data on the number of xations (Table 4) showed that both the
eects of Group (F(1, 26) 9.74, p 5 0.0044), and Type of problem-solving task (F(2,
52) 84.86, p 5 0.0001) were signicant. Post-hoc tests revealed that the number
of xations increased signicantly ( p 5 0.01) from shooting situations to dribbling
and passing situations. The dierences in xation duration between groups was not
signicant, whereas Type of problem-solving task showed a signicant main eect
(F(2, 52 39.21, p 5 0.0001). Post-hoc procedures revealed that xation duration for
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

Expertise and visual information processing

13

Table 4. Mean values and standard errors (in parentheses) of the number of xations and
xation duration (ms) for both groups in solving shotting, dribbling and passing situations,
presented by means of slides
Number of xations

Fixation duration (ms)

Group

Shooting

Dribbling

Passing

Shooting

Dribbling

Passing

Expert

2.17
(0.16)

2.67
(0.22)

4.19
(0.15)

629
(38.42)

578
(41.18)

351
(11.70)

Intermediate

2.97
(0.20)

3.53
(0.34)

4.93
(0.21)

639
(50.71)

650
(86.11)

419
(31.43)

passing situations was signicantly shorter than xation durations for both shooting
and dribbling situations (p 5 0.01).
As in other team sports, these results show that visual information processing is
dependent upon competence, with experts being more selective, and thus using fewer
xations (Bard and Fleury, 1976; Helsen and Pauwels, 1990; Helsen, Pauwels and
Van Outryve d'Ydewalle, 1986).
According to the results found in a similar basketball task (Bard and Fleury, 1976,
1981), the type of problem-solving task presented to the subjects (shooting, dribbling,
passing) signicantly modied xation duration, and overall number of xations.
These ndings suggest that items on the slides did not present the same informational
value, and led to the use of dierent strategies. Consequently, the location of xations
was studied as a function of the level of expertise, type of solution, and response
accuracy.
Fixation location
Similar results were obtained for shooting, dribbling and passing situations. For sake
of brevity only passing situations are presented (Figure 1). A 2  4  2 (Group 
Fixation location  Response accuracy) ANOVA with repeated measures on the
second and third factors was performed on trial time.
The Fixation location eect was signicant (F(3, 78) 108.56, p 5 0.0001), as
well as the Fixation location  Response accuracy interaction (F(3, 78) 44.91,
p 5 0.0001), and the overall Group  Fixation location  Response accuracy interaction (F(3, 78) 4.38, p 5 0.0067). Two-way analyses of variance were then
conducted separately for each subject group, followed by simple main eects. Both
groups focus more on the free team mate and spend less time xating the rest when
wrong decisions are made. For experts more time is spent viewing the ball possessor in
wrong decisions than in correct decisions. In addition, intermediates spend less time
xating the goal in wrong decisions. This is illustrated in Figure 1. It might be argued
that experts in this case show the same scanpath typical for shooting situations,
taking into account the relation between the player in possession of the ball and the
goal, while intermediates make more allowances for their team mates and opponents.
Generally, these data conrm previous results (Helsen and Pauwels, 1990) and
reveal rst that both experts and intermediates favour specic cues for each solution.
Second, no substantial dierences were found in the allocation of xations to display
regions as a function of player expertise. Third, the type of problem-solving task
signicantly inuenced response time, response accuracy, number of xations, and
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

14

W. F. Helsen and J. L. Starkes

Figure 1. Percentage of trial time (SE) allocated to each xation location as a function of
response accuracy in passing situations for the expert and intermediate performers

xation duration, and reected dierences in situation complexity. Shooting situations reect the simple relationship between the player in possession of the ball and
the goal, and were easier than both other problem situations.
Conclusions
The speed and accuracy results from Experiment 2 show that experts are faster and
more accurate than intermediates. The eye-movement data might suggest why this
superiority occurs. When only correct decisions are examined, there are no signicant
dierences found for xation duration or location of xations between experts and
intermediates. However, there is a signicant dierence for the number of xations.
This means that experts can do `more with less', that they can nd the best decision on
the basis of fewer xations. Experts are able to take in more relevant information
within a single xation, primarily because their vast domain-specic knowledge
allows them to chunk the present information into meaningful units, and helps them
to recall information faster and more accurately.
From the xation location data, the discrepancy in our results between selective
information pick-up and visual search (as implied from foveal orientation to the
display) was more apparent as decisions became more complex. Selectivity in a sportoriented problem-solving task appears to speed up the processing rate and result in
representation at a deeper, more principled level of processing, using abstract
principles about how the game has to be played, rather than supercial, literal features
as the basis for problem solving and knowledge structure. This is the case for elite
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

Expertise and visual information processing

15

soccer players as well as for experts in cognitive tasks (e.g. Glaser and Chi, 1988), and
sport experts in resequencing (e.g. Vickers, 1986) and categorization tasks (Allard and
Burnett, 1985).
For motor skills, it has often been argued that one requires only procedural
knowledge or knowledge of `what to do when'. Still, paradigms that tap the resources
of sport experts and do not involve strictly motor responses tell us a great deal about
the nature of declarative knowledge and its role in skilled movement. If dierences
were only found between experts and intermediates in actual games, then skill in sport
would simply reect dierences in procedural knowledge. The ndings of Experiment
2 provide further evidence that a declarative knowledge base is a valuable resource in
soccer. This knowledge exists independently of the physical attributes required in
kicking a ball. Skill in soccer reects not only procedural knowledge but also an
enhanced level of declarative knowledge and superior chunking. This knowledge
augments the motor operations of performance by producing fast and accurate
decisions within the time constraints imposed by elite-level soccer.
Yet one can hardly assess the knowledge structure of an expert player without
noting that much of the declarative base is acquired and proceduralized within the
course of movement. The contention here is that procedural knowledge, obtained in
setting up and running plays, probably enhances the acquisition and retention of
declarative knowledge. Knowing how to perform a 3 against 2 oence in soccer
provides more `hooks' on which to hang new declarative knowledge.
Several researchers (Starkes, 1987; Abernethy et al., 1994; Starkes et al., 1995) have
shown that the more task-specic and related the measured cognitive ability is to the
actual game, the better the resultant prediction of performance. Therefore, in Experiment 3, subjects were required to make the same tactical decisions as in the previous
experiment (to shoot, pass, or dribble the ball). Experiment 3 retains experimental
control and replicability and increases task delity in stimulus and response. By using
a 16 mm lm simulation task, ecological validity was enhanced for the dynamic
features of the task. In addition, a soccer-specic motor response was required instead
of a verbal response. These changes made the study unique in the literature, allowing
analysis of performance of players in the actual task soccer players do so well. In
addition to speed and decision accuracy of tactical decision making, attention was
focused on the type of visual information used and the way in which this information
was processed. The question then was: as we make the laboratory task more
perceptually and motorically realistic are expert-intermediate dierences augmented?
EXPERIMENT 3: EYE-MOVEMENT ANALYSES OF SOCCER FILM
In cognitive activities such as playing board games and solving mental problems,
think-aloud protocols have been used to demonstrate that experts consider more
alternatives and look further ahead. Experts also evaluate available options more
fully and more rapidly than novices (e.g. Holding and Reynolds, 1982). In timeconstrained activities, which require quick and accurate decision making (such as ball
sports), experts seem to focus more on anticipatory cues than novices. These cues are
available from the context or the opponent's action (Wrisberg, 1993). This has been
examined in natural settings (e.g. Howarth et al., 1984; Starkes et al., 1995), as well as
in lm simulation tasks through the use of lm-occlusion (e.g. Abernethy, 1985,
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

16

W. F. Helsen and J. L. Starkes

1991), and eye-movement registration (e.g. Helsen and Pauwels, 1993a). Until now
this approach has been applied more in individual sports, such as badminton, tennis,
and squash, and only rarely to team sports.
Method
Subjects
Again, the same subjects who participated in this experiment had also participated in
the previous experiments. These were 14 male semi-professional players and 14 male
kinesiology students. All the subjects were right-footed, as determined by a preexperimental interview and test session.
Apparatus and task
Thirty dierent tactical situations were rst selected from videotaped soccer matches
of European Cup and World Cup games (free kicks, penalty kicks, o-side problems,
dribbling, shooting, and passing situations). The nature and number of each of the
tactical actions studied on video are discussed in more detail elsewhere (Helsen and
Pauwels, 1988). These game situations were then re-enacted `live', and lmed using
16 mm lm with a camera taking the place of one of the players (see Helsen and
Pauwels, 1993b, for more information).
The simulation consisted of a lm task (Figure 2), that was projected life-size onto
a white screen set 9 m in front of the subject (10 m  4 m, 908 of visual angle). A
constant preparatory signal was given before presentation. When each situation
started, the ball moved on the screen from one player to another. At a specic moment,
the ball was played by one of the attackers in the direction of the subject who became
part of the action on the screen. In response to an auditory signal on the lm he was
required to perform, as quickly and accurately as possible, a tactical decision with the
ball just as in a real game situation. The possible responses were; shoot at the goal,

Figure 2.

The 16 mm lm simulation task (Helsen and Pauwels, 1993b)

Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

Expertise and visual information processing

17

dribble around the goalkeeper or an opponent, and pass to a specic and free team
mate. Timing was recorded automatically when the subject started his movement
(initiation time), when he touched the ball (ball/foot contact time), and when the ball
reached the screen (total response time). The times were transmitted to a microcomputer, stored on disk, and printed. The accuracy of the decision was also recorded
for each situation.
The same NAC-V was used as in Experiment 2. For the evaluation of visual
exploration in a dynamic context, however, it was important to study not only
saccades and xations but also slow-tracking eye movements. Once an object is
xated, these movements (or `foveations') keep it in central vision as it moves or as the
observer moves for a period that lasts at least 140 ms. This denition corresponds to
that used in other literature (e.g. Abernethy, 1987; Goulet et al., 1989). As in previous
work (Helsen and Pauwels, 1993b) an acceptable inter- and intra-observator
reliability was obtained (0.99 and 0.97 respectively).
Procedure
The experimental set-up was explained individually to each subject by means of a
standardized introduction on the lm. The eye-movement recording apparatus was
tted onto the subjects' head, stabilized, calibrated, and checked as in Experiment 2.
Five situations were used for familiarization, followed by 30 game situations, equally
divided in two series of fteen. Each lm sequence was separated by 10 s of black
leader lm to ensure a constant intertrial interval as well as to provide the subject
enough time to prepare for the next sequence. A distinction was also made between
two phases of response. In `the preparation phase' the team mates on the screen
realistically developed dierent attacking patterns by playing the ball from one player
to another. This part of the lm typically lasted 10 to 12 seconds. The subject was
provided with essential actions or cues, which allowed him to anticipate in the
response phase. In the second phase (`response phase'), the ball was played by one of
the attackers in the direction of the subject who became part of the action on the
screen. To trigger the response phase, an auditory signal on the lm meant he was to
perform his decision as quickly and accurately as possible.
Design
Five dependent variables were assessed. Response times (initiation time; movement
time; ball/foot contact time; ballight time; total response time) and response accuracy
were measured. To evaluate the ongoing cognitive processes, number of xations,
xation duration, and xation location were studied. Since the analyses consisted of
several separate one-way ANOVAS on data from the same subjects, only dierences
found to surpass p 5 0.002 are reported as signicant. This was done to protect against
Type I errors.
The general purpose of this study was to determine: (1) if experts were faster and
more accurate in nding tactical solutions; (2) if experts diered from less experienced
players in their selectivity and processing speed; and (3) if, according to the type of
solution, the subjects used dierent visual cues to anticipate game situations similar to
those that occur in actual games. Finally, does performance with dynamic lm stimuli
(Experiment 3) more accurately reect skill level than with static slide stimuli
(Experiment 2).
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

18

W. F. Helsen and J. L. Starkes

Table 5. Mean values and standard errors (in parentheses) of response times (ms) and
response accuracy (%) for both groups in solving game situations, presented by means of
16 mm lm
Response times (ms)
Group

Initiation Movement Ball-contact Ballight


Total
Response
time
time
time
time
response time accuracy (%)

Expert

760
(38.92)

927
(23.68)

1689
(29.44)

694
(22.13)

2363
(41.12)

92
(0.29)

Intermediate

921
(51.17)

934
(30.58)

1851
(39.12)

755
(33.19)

2604
(46.85)

82
(0.57)

Results and discussion


Response speed and accuracy
The analysis of response time data (Table 5) showed signicant dierences between
groups for initiation time (F(1, 26) 6.21, p 5 0.0194), ball/foot contact time
(F(1, 26) 10.87, p 5 0.0028), and total response time (F(1, 26) 14.94, p 5 0.0007).
Dierences in movement time (to kick the ball) and ballight time (following the
kick) were not signicant. Analysis of response accuracy showed that the main eect
of Group was signicant (F(1, 26) 20.91, p 5 0.0001). Experts were faster and were
also better able to nd the right solution.
Experts had shorter mean response times than intermediates, which supports the
hypothesis that the duration of information processing decreased as a function of
experience. Furthermore, the results for response accuracy conrm previous research
ndings in team sports that experts have more accurate decision processes than
intermediates in solving tactical game problems, presented dynamically by means
of video or 16 mm lm. (Bard and Fleury, 1981; Helsen and Pauwels, 1988, 1993a,b;
Helsen et al., 1986).
Number of xations and xation duration
The analysis of data for the number of xations showed that the dierences between
the two groups were signicant both in the preparation phase (F(1, 26) 6.62,
p 5 0.0162) and the response phase (F(1, 26) 12.33, p 5 0.0016). Intermediates were
more likely to switch attention frequently from one potential information source to
another. These ndings clearly show an increase in selectivity as a function of
increasing expertise, and conrm previous ndings in team sports (Bard and Fleury,
1981; Helsen and Pauwels, 1993a,b; Helsen et al., 1986; Ripoll, 1988). Expertise gives
rise to shorter information-processing times and, hence, shorter response times and
better performance.
The data suggest that xation duration (Table 6) for this task is substantially longer
than that generally seen in tasks where the display is static, but within the range
reported for tasks using dynamic stimuli (Abernethy and Russell, 1987). Dierences
in xation duration between groups were not signicant. No formal analysis was
conducted on what would appear to have been a systematic reduction of xation
duration as more display information became available to the subjects. We might
speculate that since xations in the preparation phase were relatively lengthy they may
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

Expertise and visual information processing

19

Table 6. Mean values and standard errors (in parentheses) of the number of xations and
xation duration (ms) for both groups in solving tactical game situations, presented by means
of 16 mm lm, in both the preparation and response phases
Number of xations
Group phase

Fixation duration (ms)

Preparation phase Response phase Preparation phase Response phase

Expert

10.78
(0.38)

1.71
(0.08)

1002
(59.99)

467
(14.00)

Intermediate

12.57
(0.58)

2.17
(0.10)

860
(47.13)

440
(17.56)

have been used for the active extraction of information of use in resolving display
uncertainty. As more information became available (in the response phase), the
search task became more one of conrming the existing information rather than
extracting new information. Presumably this conrmation can be accomplished
through the use of xations of shorter mean duration. Search rate therefore, appears
to reect quite closely task diculty and the extent of display redundancy. This is an
eect which has been hypothesized (Just and Carpenter, 1976) but often not
demonstrated in visual search studies.
Fixation location
To study what visual information is used in order to anticipate game situations,
xation location is discussed for dierent situations. This analysis, along with our
subjective observations of the search patterns, suggest that the following search
sequence was common to most subjects, irrespective of their level of expertise.
Subjects initially prepared themselves for the lm presentation by xating in
proximity to the trial number on the screen centre, thereby adopting a strategy which
not only satised the instructions provided but also provided the highest potential for
early location of the target object (i.e. the ball). Once the lm display started, there
was some inevitable latency before subjects made their rst saccadic eye movement
and this saccade was generally directed to gross sections of the scene, such as the goal
or the player in possession of the ball. It would appear that these initial xations were
primarily concerned with providing subjects with early visual information regarding
the direction of the attack, and whether the developing action was a shooting,
dribbling, or passing situation. As will be discussed for the passing situations, those
cues xated in the preparation phase were accentuated in the response phase. When
the ball was played in the direction of the subject, ball ight was very rarely sampled
exhaustively and movement of the head away from the display preceded the initiation
of the subject's action.
This observation of broken monitoring of ball ight suggests that the majority of
information conveyed by the ball in its ight is redundant, and acts only to provide
information to conrm perceptual judgements made much earlier in the attack
sequence. This concept of late ball ight redundancy is congruent with the observation of shuttle ight observation in badminton and with the observation that ocular
tracking of real ball ight is generally incomplete, and terminated some distance
before racquet or bat contact (Abernethy and Russell, 1987; Lyons, Fontaine and
Elliott, 1997). In the instances where ball ight was maintained in the early stages of
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

20

W. F. Helsen and J. L. Starkes

Figure 3. Percentage of trial time (SE) allocated to each xation location in the preparation
phase of passing situations for the expert and intermediate performers

ight it was done through the use of saccadic rather than smooth pursuit (or tracking)
eye movements. It also questions those studies which have attempted to dierentiate expert and novice performers on the basis of simple ocular tracking tasks
(Trachtman, 1973), or enhance skill acquisition through training the eyehead system
alone (e.g. McLeod, 1991).
In Figure 3, xation location in the preparation phase is shown for the passing
situations. A 2  7 (Group  Fixation location) ANOVA with repeated measures on
the last factor was performed on trial time. A signicant main eect for Fixation
location (F(6, 132) 17.70, p 5 0.0001) was found. Post-hoc tests revealed a signicant dierence between xations on the attackers versus all other categories. Another
signicant dierence occurred between xations on the free team mate versus the ball
(p 5 0.01), and the free space ( p 5 0.01).
A similar 2  7 (Group  Fixation location) ANOVA on trial time in the response
phase (Figure 4) showed a signicant main eect for Fixation location (F(6, 132)
13.88, p 5 0.0001). Post-hoc tests indicated signicant dierences between xations on
the attacking team mates and the free team mate versus all other categories (p 5 0.01).
From both Figures 3 and 4 it is clear that the small dierences in the preparation
phase, were augmented in the response phase. These data conrm previous soccer
xations ndings (Helsen and Pauwels, 1993b).
With regard to the number of xations and their duration, the results found in team
sports as discussed here and elsewhere (Bard and Fleury, 1976; Helsen et al., 1986)
dier from those obtained in individual sports, where experts showed similar scan
patterns as novices (e.g. Abernethy and Russell, 1987). However, as Goulet et al.
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

Expertise and visual information processing

21

Figure 4. Percentage of trial time (SE) allocated to each xation location in the response
phase of passing situations for the expert and intermediate performers

(1989) have noted, the task requirements in each are obviously dierent. In the team
sport studies conducted thus far, subjects were allowed to control the time of
exposure, and the situation as a whole was presented at the beginning of the slide
presentation. In individual sport studies to date, the exposure time is similar to all
subjects and information presentation is not completed until the end of the lm
presentation. According to Steiner's (1966), typology in the former case we are
dealing with a maximizing task, in which time requirements are emphasized; the
shorter, the better. In the latter case, an optimizing task, based on the quality of
the response, is presented. Since experts process information more rapidly, they
analyse a greater amount of information during the allocated time in an optimizing
task and consequently have increased performance. Furthermore, in many of these
lm tasks no time constraints are imposed on decision making, encouraging subjects
to use far more predictable orders of search than when either the display is static
and/or the search task is time constrained. In a static display all sections of the display
may be equally informative whereas in a dynamic situation informativeness is
restricted primarily to those spatial regions which contain features changing as a
function of time. Finally, the tasks presented in individual sports did not involve a
sport-specic response. According to Ripoll (1991), the dierences in the results
found in team sports and in individual sports, are not so unequivocal as Abernethy
and Russell (1987) suggest, but rather a consequence of dierences in the nature of the
task.
As we can infer both from the results of the movement response and eye-movement
evaluation, expert players extract information eectively prior to the response signal
to anticipate and thereby reduce initiation time. This is also shown by the strong
relationship (r 0.75) between initiation time and total response time.
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

22

W. F. Helsen and J. L. Starkes

Conclusions
From Experiment 3 evidence was found that in a dynamic, contextually relevant
task experts were able to extract early information prior to the response signal. This
was not true of lesser-skilled subjects. Apart from allowing the expert performer to
respond earlier, this anticipation may also eectively reduce the expert's informationprocessing load. Superior anticipatory skill may explain the more rapid response
times frequently reported for experts performing sport-specic decision-making
tasks in both laboratory (e.g. Abernethy, 1991; Abernethy and Russell, 1984, 1987)
and natural settings (e.g. Howarth et al., 1984; Starkes et al., 1995). Furthermore, it
provides evidence of the dierential cue availability for dierent oensive plays used
in soccer.
In the transition from a cognitive tactical decision to a motor response, soccer
experts make better use of the information aorded from defenders in general, and the
free back, and free space in particular. Intermediate players look primarily at the ball,
the attackers, and the goal. That experts make better use of the position game of the
free back or sweeper perhaps can be explained by the enhanced role of this player
position in recent years. He must not only act as the organizer of the defense but also
provide the attack impulse as the rst and undefended attacker from the second line.
We would speculate that intermediate players cannot assess the informational value of
the free back position. In the same way, less-skilled players probably make much less
use of the errors that are made by the defence, and of the information available in the
free space. This has also been noted by other researchers (Bard and Fleury, 1976,
1981).
A larger database of declarative knowledge, however, does not always guarantee
better performance. Hence, in the second instance, experts dier in the eciency with
which they can link environmental information to declarative knowledge. The key to
this lies in the experts' ability to chunk in meaningful units a larger amount of
structured information within one single xation.
In some ways the expert/intermediate dierences seen in these studies may even
underestimate the degree of dierence in the real world. A discussion of the responses
to the lm task will serve as an example. Recall that in this experiment subjects saw a
game situation evolve on lm. In response to an auditory signal, the subject was
instructed to respond as quickly and accurately as possible. Dependent measures were
response speed and accuracy, as well as a number of xations, xation duration, and
xation location. The methodology in this case ensured that a standardized amount
of information was seen by both experts and intermediates. The various dependent
measures assessed how and where level of expertise impacted decisions. It is possible
that experts in a real game situation would make decisions earlier, and based on less
information than provided before the response signal. Experts may be capable of
using advance visual cues or probabilities to make a much earlier decision. If this were
the case then beyond a certain point eye xations may simply have been used to
conrm earlier decisions. One way around this confound would be to allow players to
view only as much lm as they felt necessary to make a decision. In some ways then,
the large amount of performance variability explained by cognitive aspects of
expertise may still have been underestimated.

Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

Expertise and visual information processing

23

A MULTIDIMENSIONAL APPROACH TO PREDICTING PERFORMANCE


Method
Recall that the primary goal of these studies was to provide a multidimensional
approach to skilled perception. Wrisberg (1993) has emphasized that expertise
research needs to be interactionist and multidimensional, and examined the grouping
of factors that contribute most to skilful performance in particular domains. For the
rst time it is possible to assess the relative contribution to both non-specic abilities
and domain-specic skills to skill in soccer, and to compare these data with the results
found in an equally highly structured team sport, eld hockey (Starkes and Deakin,
1984; Starkes, 1987). A stepwise discriminant analysis was carried out, using all the
variables discussed in the previous three experiments (Table 1).
Results and discussion
The results indicated an average squared canonical correlation of 0.84 with the
signicant step variables all being domain-specic variables. The primary variable
was response accuracy (Film; F 20.91, r2 0.45, p 5 0.0001). The second and third
variables were the number of xations in solving tactical game problems, presented
dynamically by means of lm (response phase; F 18.12, r2 0.42, p 5 0.0003) and
statically by means of slides (F 7.61, r2 0.24, p 5 0.0109). The fourth variable was
the number of xations in solving tactical game problems, presented dynamically by
means of lm (preparation phase; F 5.12, r2 0.18, p 5 0.0333). The fth variable
was ball-contact time (F 5.19, r2 0.19, p 5 0.0328). Generally, these variables can
be divided into perceptual (number of xations), decision (response accuracy), and
motor (ball-contact time) components. Together these variables explained 84% of the
variance in performance of both skill groups. The last and only non-specic variable
was peripheral visual range in the horizontal dimension (F 5.56, r2 0.20,
p 5 0.0277), with a small contribution of 3%.
The results of this multidimensional approach support the ndings of Starkes and
Deakin (1984), and Starkes (1987) for eld hockey. To predict group membership
using nine variables, their results indicated a multiple R 0.83 (r2 0.69) with two
signicant step variables, both being domain-specic variables. The primary variable
was recall of game-structured information and the second variable shot prediction
accuracy following view of ball impact. No other variables contributed signicantly in
the prediction of group membership for expert, intermediate or novice players.
The present studies again reveal the importance of the cognitive dimension of visual
search in soccer, and for the rst time suggest how much of the variance in performance can be explained by these factors.
Conclusions
The results clearly showed that superior skill is attributable to a variety of processes.
In combination with more pertinent selection and more accurate interpretation of
environmental cues (i.e. perceptual component) and more rapid selection of an
appropriate response (i.e. decision component), the advanced soccer player is able to
execute a smooth and ecient movement (i.e. motor component). These ndings
conrm, as McPherson and Thomas (1989) and Allard and Starkes (1991) have noted,
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

24

W. F. Helsen and J. L. Starkes

that a distinguishing feature of experts is their adeptness at both `knowing' what to do


and `doing it'. While less-skilled persons may achieve a degree of success with one or
the other of these capabilities, they are unable to `link' the two (e.g. an experienced
substitute player who completely understands the game but is unable to produce the
appropriate movements skilfully or the player who demonstrates skilful movements in
practice but produces them at inappropriate moments in actual competition). In
addition to these ideas, Allard and Starkes (1993) stressed that one-way automaticity
or expertise might be assessed is to determine the degree to which a performer has
achieved an appropriate `linkage' between knowing and doing (i.e. exibility in linking
rather than the establishment of stable conditionaction links). In any case, and
according to Wrisberg (1993), this concept of linkage-development is an appealing one
that deserves the attention of future investigators of motor expertise.
Although this study reveals a substantial amount about the nature of skill and the
structure of knowledge in elite soccer players, it also reveals as much about the nature
and cognitive requirements of the game. Many athletes, coaches, researchers, as well
as people in general, still misunderstand and underappreciate the importance of the
cognitive demands in dynamic sport settings. In team sports where the environment is
constantly changing decisions and responses have to be made quickly and accurately.
From studies using the multidimensional approach, several trends are beginning to
emerge. First, in the prediction of expertise the proportion of variance accounted for
by domain-specic skills seem to be substantially higher than for non-specic
abilities. Second, many of the cognitive characteristics of expert performers seen in
passive games such as chess and bridge also emerge as characteristic of highly skilled
athletes. This is apparent from the critical interaction of skill and game structure
information on the slide and lm projection tasks. While the motor component is a
given in any level of performance in soccer, there are now obvious and important
perceptual and decision components to that performance. Third, the more taskspecic and related the measured cognitive ability is to actual game requirements, the
better the resultant prediction of performance. The results of the multivariate
analyses stress the importance of imposing time constraints on decision making and
providing a linkage between knowing and doing in lm simulation tasks.
ACKNOWLEDGEMENTS
The experiments were funded in part through a Research Council Grant from the
Katholieke Universiteit Leuven. The authors wish to thank Dr Fran Allard,
Dr Martinus Buekers, Dr Jim Cauraugh, and Dr Anders Ericsson for reviewing an
earlier draft of this manuscript; Mr Pol Meugens, Mr Mark Beirinckx, and Mr Mathy
Vanbuel for providing invaluable guidance in designing the research equipment and
the electronics. Requests for reprints should be sent to Dr Werner Helsen, Leuven
University, Motor Learning Laboratory, Tervuursevest 101, 3001 Leuven, Belgium.
REFERENCES
Abernethy, B. (1985). Cue usage in `open' motor skills: a review of the available procedures.
In D. G. Russell and B. Abernethy (Eds), Motor memory and control: the Otago symposium.
(pp. 110122). Dunedin, New Zealand: Human Performance Associates.
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

Expertise and visual information processing

25

Abernethy, B. (1987). Anticipation in sport: a review. Physical Education Review, 10, 516.
Abernethy, B. (1991). Visual search strategies and decision-making in sport. International
Journal of Sport Psychology, 22, 189210.
Abernethy, B., Neal, R. J. and Koning, P. (1994). Visual-perceptual and cognitive dierences
between expert, intermediate and novice snooker players. Applied Cognitive Psychology, 8,
185211.
Abernethy, B. and Russell, D. G. (1984). Advance in cue utilisation by skilled cricket batsmen.
Australian Journal of Science and Medicine in Sport, 16, 210.
Abernethy, B. and Russell, D. G. (1987). The relationship between expertise and visual search
strategy in a racquet sport. Human Movement Science, 6, 283319.
Allard, F. and Burnett, N. (1985). Skill in sport. Canadian Journal of Psychology, 39, 294312.
Allard, F., Deakin, J., Parker, S. and Rodgers, W. (1993). Declarative knowledge in skilled
motor performance: byproduct or constituent? In J. L. Starkes and F. Allard (Eds),
Cognitive issues in motor expertise (pp. 95107), Amsterdam: North-Holland.
Allard, F., Graham, S. and Paarsalu, M. E. (1980). Perception in sport: basketball. Journal of
Sport Psychology, 2, 1421.
Allard, F. and Starkes, J. L. (1980). Perception in sport: volleyball. Journal of Sport
Psychology, 2, 2223.
Allard, F. and Starkes, J. L. (1991). In K. A. Ericsson and J. Smith (Eds), Toward a general
theory of expertise: Prospects and limits (pp. 126152). Cambridge: Cambridge University
Press.
Bard, C. and Fleury, M. (1976). Analysis of visual search activity during sport problem
situations. Journal of Human Movement Studies, 3, 214222.
Bard, C. and Fleury, M. (1981). Considering eye movements as a predictor of attainment.
In I. M. Cockerill and W. W. McGillivary (Eds), Vision and sport (pp. 2841). Cheltenham:
Stanley Thornes.
Chase, W. G. and Simon, H. A. (1973). Perception in chess. Cognitive Psychology, 4, 5581.
Doil, W. and Binding, M. (1986). Peripheres Sehen als Voraussetzung fur die Orientierung in
Sportspielen. Medicin und Sport, 26, 5558.
Ericsson, K. A. (Ed.) (1996). The road to excellence: The acquisition of expert performance in
the Arts and Sciences, Sports and Games. Hillsdale, NJ: Erlbaum.
Ericsson, K. A. and Lehmann, A. C. (1996). Expert and exceptional performance: evidence of
maximal adaptation to task constraints. Annual Review of Psychology, 47, 273305.
Fleishman, E. A. (1966). On the relationship between abilities, learning, and human
performance. American Psychologist, 27, 10171032.
French, K. E. and Thomas, J. R. (1987). The relation of knowledge development to children's
basketball performance. Journal of Sport Psychology, 9, 1532.
Gangemi, P. F., Zaccara, G., Bordiga, M., Caldarone, G., Minneo, L., Messori, A., Parigi, A.
and Luzzi, S. (1993). Oculomotor and manual tracking in a group of Olympic gymnasts.
In G. d'Ydewalle and J. Van Resenbergen (Eds), Perception and cognition: Advances in eye
movement research (pp. 149157), Amsterdam: North-Holland.
Glaser, R. (1996). Changing the agency for learning: Acquiring expert performance. In K. A.
Ericsson (Ed.), The road to excellence: The acquisition of expert performance in the Arts and
Sciences, Sports and Games (pp. 303311). Hillsdale, NJ: Erlbaum.
Glaser, R., Chi, M. T. H. (1988). Overview. In M. T. H. Chi, R. Glaser and M. J. Farr (Eds),
The nature of expertise (pp. 1528). Hillsdale, NJ: Erlbaum.
Goulet, C., Bard, C. and Fleury, M. (1989). Expertise dierences in preparing to return a
tennis serve: a visual information processing approach. Journal of Sport and Exercise
Psychology, 11, 382398.
Helsen, W. and Pauwels, J. M. (1988). The use of a simulator in evaluation and training of
tactical skills in football. In T. Reilly, A. Lees, K. Davids and W. J. Murphy (Eds), Science
and football (pp. 493497). London: Spon.
Helsen, W. and Pauwels, J. M. (1990). Analysis of visual search activity during tactical
game problems. In D. Brogan (Ed.), Visual search I (pp. 177184). London: Taylor &
Francis.
Helsen, W. and Pauwels, J. M. (1993a). A cognitive approach to visual search in sport.
In D. Brogan, A. Gale and K. Carr K (Eds), Visual search II (pp. 379388).
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

26

W. F. Helsen and J. L. Starkes

Helsen, W. and Pauwels, J. M. (1993b). The relationship between expertise and visual
information processing in sport. In J. Starkes and F. Allard (Eds), Cognitive issues in motor
expertise (pp. 109134). Amsterdam: North-Holland.
Helsen, W., Pauwels, J. M. and Van Outryve d'Ydewalle, G. (1986). Analysis of visual search
patterns during sports games. In R. Andresen (Ed.), Beitrage zur Sportspielforschung
(pp. 6981). Ahrenburg: Czwalina.
Helsen, W. F., Starkes, J. L., Elliott, D. and Ricker, K. L. (in press). Sampling frequency and
the study of eyehand coordination in aiming. Behavior Research Methods, Instruments, &
Computers.
Holding, D. H. and Reynolds, J. R. (1982). Recall or evaluation of chess positions as
determinants of chess skill. Memory and Cognition, 10, 237242.
Howarth, C., Walsh, W. D., Abernethy, B. and Snijder, W. (1984). A eld examination of
anticipation in squash: some preliminary data. The Australian Journal of Science and
Medicine in Sport, 16, 711.
Howe, M. J. A., Davidson, J. W. and Sloboda, J. A. (in press). Innate talents: reality or myth?
Behavioural and Brain Sciences.
Just, M. A. and Carpenter, P. A. (1976). Eye xations and cognitive processes. Cognitive
Psychology, 8, 441480.
Keele, S. W. and Hawkins, H. L. (1982). Explorations of individual dierences relevant to high
level skill. Journal of Motor Behavior, 14, 323.
Keele, S. W., Ivry, R. I. and Pokarny, R. A. (1987). Force control and its relation to timing.
Journal of Motor Behavior, 19, 96114.
Lyons, J., Fontaine, R. and Elliott, D. (1997). I lost it in the lights: The eects of predictable
and unpredictable intermittent vision on unimanual catching. Journal of Motor Behavior, 29,
113118.
McLeod, P. (1987). Visual reaction time and high-speed ballgames. Perception, 16, 4959.
McLeod, P. (1991). Eects of eyerobics visual skills training on selected performance measures
of female varsity soccer players. Perceptual and Motor Skills, 72, 863866.
McPherson, S. L. (1993). Knowledge representation and decision-making in sport. In J. Starkes
and F. Allard (Eds), Cognitive issues in motor expertise (pp. 159188). Amsterdam: NorthHolland.
McPherson, S. L. and Thomas, J. R. (1989). Relation of knowledge and decisions in boys'
tennis: Age and expertise. Journal of Experimental Child Psychology, 48, 190211.
Mizusawa, K., Sweeting, R. L. and Knouse, S. B. (1983). Comparative studies of color elds,
visual acuity elds, and movement perception limits among varsity athletes and non-varsity
groups. Perceptual and Motor Skills, 56, 887892.
Morris Don, G. S. (1977). Dynamic visual acuity: implications for the physical educator and
coach. Motor skills: theory into practice, 2, 1520.
Morris Don, G. S. and Kreighbaum, E. (1977). Dynamic visual acuity of varsity women
volleyball and basketball players. Research Quarterly, 48, 480483.
Pauwels, J. M. and Helsen, W. (1986). Relationship of static and dynamic displays of game
problems. In R. Andresen (Ed.), Beitrage zur Sportspielforschung (pp. 9099). Ahrenburg:
Czwalina.
Ripoll, H. (1988). Analysis of visual scanning patterns of volleyball players in a problem
solving task. International Journal of Sport Psychology, 19, 925.
Ripoll, H. (1991). The understandingacting process in sport: The relationship between the
semantic and the sensorimotor visual function. International Journal of Sport Psychology,
22, 221243.
Rutt Leas, R. and Chi, M. T. H. (1993). Analyzing diagnostic expertise of competitive
swimming coaches. In J. Starkes and F. Allard (Eds), Cognitive issues in motor expertise
(pp. 7594). Amsterdam: North-Holland.
Sanderson, F. H. (1981). Visual acuity and sports performance. In I. M. Cockerill and
W. W. MacGillivary (Eds), Vision and sport (pp. 6479). Cheltenham: Stanley Thornes.
Sanderson, F. H. and Whiting, H. T. A. (1978). Dynamic visual acuity: a possible factor in
catching performance. Journal of Motor Behavior, 10, 714.
Schneider, W. and Bjorklund, D. F. (1992). Expertise, aptitude and strategic remembering.
Child Development, 63, 461473.
Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

Expertise and visual information processing

27

Schutz, R. W. and Gessaroli, M. E. (1987). The analysis of repeated measures designs


involving multiple dependent variables. Research Quarterly for Exercise and Sport, 58,
132149.
Starkes, J. L. (1987). Skill in eld hockey: The nature of the cognitive advantage. Journal of
Sport Psychology, 9, 146160.
Starkes, J. and Allard, F. (Eds) (1993). Cognitive issues in motor expertise. Amsterdam: NorthHolland.
Starkes, J. L., Allard, F., Lindley, S. and O'Reilly, K. (1994). Abilities and skill in basketball.
International Journal of Sport Psychology, 25, 249265.
Starkes, J. L. and Deakin. J. (1984). Perception in sport: a cognitive approach to skilled
performance. In W. F. Straub and J. M. Williams (Eds), Cognitive sport psychology
(pp. 115128). Lansing: Sport Science Association.
Starkes, J. L., Edwards, P., Dissanayake, P. and Dunn, T. (1995). A new technology and eld
test of advance cue usage in volleyball, Research Quarterly, 65, 16.
Steiner, I. D. (1966). Models for inferring relationships between group size and potential group
productivity. Behavioral Sciences, 11, 273283.
Trachtman, J. N. (1973). The relationship between ocular motilities and batting averages in
little leaguers. American Journal of Optometry and Archives of the American Academy of
Optometry, 50, 914919.
Vickers, J. N. (1986). The resequencing task: determining expertnovice dierences in the
organization of a movement sequence. Research Quarterly for Exercise and Sport, 57,
260264.
Williams, M. and Davids, K. (1995). Declarative knowledge in sport: A byproduct of
experience or a characteristic of expertise. Journal of Sport and Exercise Psychology, 17,
259278.
Williams, J. M. and Thirer, J. (1975). Vertical and horizontal peripheral vision in male and
female athletes and nonathletes. Research Quarterly, 46, 200205.
Wright, D. L., Pleasants, F. and Gomez-Meza, M. (1990). Use of advanced cue sources in
volleyball. Journal of Sport and Exercise Psychology, 12, 406414.
Wrisberg, C. A. (1993). Levels of performance skill. In R. Singer, M. Murphy and K. Tennant
(Eds), Handbook of sport psychology (pp. 6172). New York: Macmillan.
Yandell, K. M. and Spirduso, W. W. (1981). Sex and athletic status as factors in reaction
latency and movement time. Research Quarterly for Exercise and Sport, 52, 495504.

Copyright # 1999 John Wiley & Sons, Ltd.

Appl. Cognit. Psychol. 13: 127 (1999)

You might also like