You are on page 1of 23

Sport Psychology

Journal of Sport & Exercise Psychology, 2007, 29, 147-169 


© 2007 Human Kinetics, Inc.

The Effects of Task Constraints on Visual


Search Behavior and Decision-Making
Skill in Youth Soccer Players
Roel Vaeyens,1 Matthieu Lenoir,1 A. Mark Williams,2
Liesbeth Mazyn,1 and Renaat M. Philippaerts1
1
Ghent University and 2Liverpool John Moores University

We examined differences in visual search behaviors and decision-making skill


across different microstates of offensive play in soccer using youth participants
(13.0–15.8 years) varying in skill and experience. We used realistic film simula-
tions of offensive play, movement-based response measures, and an eye movement
registration technique. Playing experience, skill level, and the unique constraints
of the task, expressed by the number of players and relative proportion of offen-
sive and defensive players, determined both the observed search behavior and
processing requirements imposed on players in dynamic offensive team simula-
tions. Significant differences in performance were observed between players and
nonplayers and across three groups of soccer players who differed in skill level.
Implications for talent identification and development are considered.

Key Words: mediating mechanisms, perceptual-cognitive skill, performance, task


constraints

The mechanisms underlying expert performance have aroused much interest


in recent decades (e.g., see Starkes, Helsen, & Jack, 2001; Williams & Ericsson,
2005). In particular, there is increasing awareness that skilled perception precedes
and determines appropriate action in sport (Savelsbergh, Williams, Van der Kamp, &
Ward, 2002). Athletes, coaches, and scientists have argued that perceptual-cognitive
skills play a crucial role in sports performance, particularly at the highest levels.
It is widely accepted that skilled performers are not endowed with superior visual
skills per se, but rather their enhanced cognitive knowledge bases enable them
to pick up and interpret perceptual information more effectively than less skilled
counterparts (Helsen & Starkes, 1999; Williams & Grant, 1999).
The majority of researchers have focused on the performer’s ability to antici-
pate future action requirements. This ability is often regarded as one of the most
important perceptual-cognitive skills underlying effective performance (Williams,

Vaeyens, Lenoir, Mazyn, and Philippaerts are with the Faculty of Medicine and Health Sciences, Depart-
ment of Movement and Sports Sciences, Gent, Belgium, and A. Mark Williams is with the Research
Institute for Sport and Exercise Sciences, Liverpool John Moores University, Liverpool, England.

    147
148   Vaeyens, Lenoir, Williams, Mazyn, and Philippaerts

Ward, Knowles, & Smeeton, 2002). In relation to team ball games such as soccer,
the consensus is that, when compared with less skilled counterparts, skilled ­players
(a) are faster and more accurate in recognizing and recalling patterns of play (Helsen
& Pauwels, 1993; Williams, Hodges, North, & Barton, 2006; ­Williams & Davids,
1995), (b) are superior in picking up contextual cues based on an opponent’s
postural orientation (Williams & Burwitz, 1993), (c) have enhanced knowledge
of situational probabilities (Williams, 2000; Ward & Williams, 2003), and (d) dis-
play more efficient and effective visual search strategies (Helsen & Starkes, 1999;
­Williams & Davids, 1998).
Although the task of selecting the next best move has been suggested to be
the most successful method of evaluating expert performance (De Groot, 1978),
there has been relatively limited research on the strategies employed by athletes
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

when making such decisions, especially in soccer (for exceptions, see Helsen &
Pauwels, 1993, Helsen & Starkes, 1999; Ward & Williams, 2003). Helsen and col-
leagues examined differences in tactical decision-making skill between experienced
and intermediate-level adult soccer players. Participants watched offensive open and
closed (i.e., set plays) soccer simulations and were asked to respond quickly and
accurately. The experienced players were faster and more accurate in ­responding
and employed fewer fixations of shorter duration when compared with their less
skilled counterparts. Ward and Williams (2003) assessed the role of expectations
(i.e., situational probabilities) in decision making by requiring youth soccer ­players
to highlight key players in a good position to receive the ball after viewing filmed
sequences taken from competitive matches. The results revealed that the elite
soccer players demonstrated superior perceptual-cognitive skills compared to their
sub-elite counterparts, with these differences being evident as early as 9 years of
age. These complex adaptations in knowledge structures as a result of extensive
task-specific practice are an important precursor to skilled decision making (French
& McPherson, 1999).
Thus far, few researchers have systematically examined the important mecha-
nisms underlying decision-making skill. This issue may be addressed by examining
the visual search behaviors employed as performers attempt to make such decisions.
A performer’s visual search strategy is typically assessed using an eye movement
recording system. Such a system measures both the orientation of the fovea and the
nature of the eye movements that separate each visual fixation (Williams, Janelle,
& Davids, 2004). Although point of gaze may not always be directly related to
visual attention and information extraction (e.g., for a detailed discussion of
potential limitations, see Williams et al., 2004; Zelinsky, Rao, Hayhoe, & Ballard,
1997), the recording of eye movement behavior helps identify the manner in which
performers scan the display while performing a particular task (see Williams &
Ericsson, 2005). A performer’s point of gaze is typically moved to another location
within the display using a visual saccade. However, these saccades are considered
relatively inactive periods of information processing (Wright & Ward, 1994) such
that theoretically it is assumed that a more efficient search strategy is characterized
by relatively fewer fixations of long duration (Abernethy, 1990; Williams, Davids,
& Williams, 1999). Nevertheless, this hypothesis has not always been supported
in the literature.
The inconsistency may arise because soccer players change the search strategy
employed as a function of the unique constraints presented by the task. For example,
Visual Search Behavior and Decision Making    149

in the case of central defensive players in soccer, when the ball is in the other half
of the field (as personified by 11 vs. 11 film simulations), they have to be aware
of many sources of perceptual information, located disparately across a large area
of the field. An extensive search strategy involving many fixations of relatively
short duration is required in order to make themselves aware of the positions and
movements of players, and the passing opportunities presented to the player in
possession of the ball (Williams, Davids, Burwitz, & Williams, 1994). Conversely,
when the ball is nearer the goal, such as in situations involving 3 vs. 3 on the edge
of the penalty area, lower search rates seem preferable owing to the increased role
of peripheral vision (Williams & Davids, 1998). In these more time-constrained
situations, there is empirical evidence to suggest that players fixate gaze centrally,
typically on the ball or player in possession of the ball, and use peripheral vision
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

to monitor movements “off the ball.” The use of these so-called visual pivots has
also been reported in other sports (e.g., see Ripoll, 1991; Williams & Elliott, 1999).
Finally, in a typical “dual” situation, as personified by 1 vs. 1 simulations, fewer
fixations of longer duration are once again employed compared to the 11 vs. 11
scenarios because players need to extract information from relatively fewer areas of
interest (i.e., only the opponent’s postural orientation and movements). Similarly,
­Martell and Vickers (2004) showed that expert hockey players changed their search
strategies when confronted with different playing environments. The observed
differences in search pattern were attributable to the task constraints imposed as
determined by the size of the playing area, the number of players involved, and
the player’s intentions.
A suggestion is that the search rate is dependent on the degree of nonfoveal
retinal stimulation, with the latter being somewhat dependent on image size (Harris,
1989; Harris, Hainline, Abramov, Lemerise, & Camenzuli, 1988). As retinal image
size increases, more nonfoveal retina is stimulated, which raises the probability of
a saccade being triggered, thus increasing the number of fixations and decreasing
average fixation duration. However, the size of the image is also likely to interact
with the number of information sources present within the display (e.g., number
of players). For example, in the context of defensive play in soccer, using film
simulations assuming that image or screen size remains constant, when only a
few players are presented in the display a search strategy involving few fixations
of long duration may be needed to extract pertinent information from only a few
key players (Williams & Davids, 1998). As the number of players increases, the
extensive number of perceptual information sources, located disparately across a
larger area of the field, increases the difficulty of using peripheral vision to extract
information, thus constraining the experienced observer to utilize more frequent
fixations of shorter duration (Williams et al., 1994).
To date, no systematic research has been undertaken to analyze whether differ-
ences in search behaviors exist across the various microstates of offensive play in
team ball games. Williams (2000) concluded that the nature of the task constrains
skilled players to use different visual search strategies during offensive (see Helsen
& Starkes, 1999) and defensive (see Williams et al., 1994; Williams & Davids,
1998) situations. Specifically, defensive situations require players to employ a more
exhaustive search pattern involving a higher number of fixations of shorter duration
than that exhibited in comparable offensive situations. The player’s intentions and
strategic role within the team guide the search for information. The key difference
150   Vaeyens, Lenoir, Williams, Mazyn, and Philippaerts

is that defensive players have to be aware of all potential passing options facing
the player in possession of the ball so as to plan ahead and prevent a goal-scoring
opportunity, whereas, in contrast, the offensive player has more control over the
course of action to be taken and may select a suitable passing option just as soon as
an opportunity arises (Williams et al., 2004). Such an issue warrants investigative
effort in order to facilitate our understanding of how players extract information
from the display in situations involving dynamic decision making and to help
develop appropriate perceptual-cognitive training programs (e.g., see Williams et
al., 2002; Smeeton, Williams, Hodges, & Ward, 2005).
The majority of researchers have employed the expert–novice paradigm to
isolate the important characteristics that differentiate skilled from less skilled
individuals (Williams et al., 2002). Generally, researchers have examined two
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

groups of athletes (e.g., soccer players) that are differentiated both in relation to
experience and playing level. The apparent confound created by the variation in
participant skill and experience level has proved problematic, particularly when
attempting to determine the relative contribution of innate talent and experience
on the task in fostering expert performance. The difficulty is compounded by the
apparent absence of adequate control groups differentiated on the skill and expe-
rience continuum (Abernethy, Thomas, & Thomas, 1993; for an exception, see
Williams & Davids, 1995). An important question is whether existing paradigms
are able to discriminate effectively among players of varying skill level, but with
a comparable amount of soccer experience (Williams et al., 1999). Moreover, the
majority of researchers have relied almost exclusively on adult sample groups (for
recent exceptions, see Reilly, Williams, Nevill, & Franks, 2000; Ward & Williams,
2003). Because the quality of the practice and match play experience and the level
of instruction provided by significant others may equally be responsible for the dif-
ferences in decision-making skill across groups, an interesting undertaking would be
to investigate differences between youth players of various competitive levels with
equivalent years of playing experience in relation to their decision-making skill.
In the current study we attempted to address the perceived shortcomings in
the literature highlighted above. We examined differences in decision-making skill
and visual search strategies across five categories of small-sided, offensive game
simulations in soccer (2 vs. 1, 3 vs. 1, 3 vs. 2, 4 vs. 3, and 5 vs. 3) using adoles-
cent participants varying in playing experience and skill level. The simulations
represent some of the typical situations that occur during real competitive games.
We present the first attempt to systematically manipulate the number of players
involved in open play offensive simulations using the same group of participants
and methods, and as such the data are partly exploratory. However, based on
published literature involving various defensive simulations, we considered that
the specific nature of the task determines both the visual search behaviors and
processing requirements imposed on players in dynamic team situations. More
specifically, we hypothesized that the search rate employed by participants would
vary with the increase in processing load (i.e., more players and a higher relative
amount of defenders) and potentially as a result of the relative importance of the
fovea, parafovea, and visual periphery in extracting information from the display.
Similarly, it was hypothesized that participants’ accuracy and response times would
differ across the various microstates of the game. We presumed that an increase in
the number of players and a lower ratio between offensive and defensive players
Visual Search Behavior and Decision Making    151

would increase the complexity of the task, providing more information to process
and increasing decision time and/or reducing response accuracy.
A secondary aim was to determine key differences in decision-making skill
and visual search behaviors using youth participants varying in skill and experi-
ence. We employed three groups of soccer players with different skill levels, but
comparable amounts of experience, and a group of nonplayers. In agreement with
previous results, we hypothesized that more refined task-specific knowledge struc-
tures would enable youth soccer players to demonstrate superior decision-making
skill and more pertinent visual search behaviors compared with nonplayers. We
also predicted that differences would be observed across the three groups of elite,
sub-elite, and regional players, although these players were reasonably well matched
in relation to the amount of accumulated experience. Specifically, we hypothesized
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

that the decision-making skill would increase concomitantly with playing level. The
elite players were expected to have developed more refined knowledge structures,
thereby facilitating effective decision making, as characterized by faster decision
times and/or greater response accuracy.

Methods
Participants
A total of 87 male adolescents (13.0–15.8 years) were recruited. Participants
were assigned to one of four subgroups according to their experience and playing
level. The elite group consisted of 21 youth players (M age = 14.7 ± 0.5 years)
who had been playing soccer for 8.5 ± 1.4 years, averaging 427 games per player.
These youngsters were recruited from soccer academies in Belgium and played
in the youth team of one of the first-division clubs (i.e., the highest level). The
majority of elite participants had represented the national youth team in Belgium.
The sub-elite group included 21 (M age = 14.6 ± 0.3 years) soccer players from
second- and third-division teams. The sub-elite players had been playing soccer
for 8.2 ± 1.1 years with an average of 404 matches per player. Half the players had
represented a provincial youth team in the current or previous competitive season.
The majority of participants from the elite and sub-elite groups played in the same
youth competition. The third subgroup consisted of 23 soccer players (M age =
14.6 ± 0.6 years) who played in regional teams with an average of 7.3 ± 1.4 years
of experience and 298 matches per player. The control group included 22 students
(M age = 14.5 ± 0.4 years) who had not participated in team ball sports during the
last 5 years. The latter participants had limited or no soccer experience. Informed
(parental) consent was obtained prior to participation. Ethical approval was provided
through the Ghent University Hospital.

Test Film
Participants were presented with offensive patterns of play. These were filmed by
a professional cinematographer from an elevated perspective (approx. 3 m) in the
center circle area behind one of the attackers. This filming perspective provided
a wide viewing angle and some perspective so as to facilitate the perception of
depth. The viewing angle provided the closest correspondence with the field of
152   Vaeyens, Lenoir, Williams, Mazyn, and Philippaerts

view typically observed by a central offensive player during a match, covering most
of the offensive half of the pitch. After detailed instruction and rehearsal over a
2‑day period, skilled (elite) offensive and defensive peer-aged youth soccer players
realistically simulated the offensive patterns. The defenders wore a black playing
outfit, whereas the attackers were dressed in red. Participants were required to
imagine themselves as the offensive player positioned just in front of the camera.
In order to clearly identify this player, the player wore a yellow vest over the red
playing shirt. This player was referred to as the yellow player. After extensive pilot
work, the decision was made to include the yellow player rather than to film from
a “first-person” perspective, in which the camera takes the place of this player. The
latter approach provided a relatively narrow field of view, even with a wide-angle
lens, and made it extremely difficult to capture the action footage in question,
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

particularly in sequences of play involving more than five players (cf. Williams
& Davids, 1998; Williams et al., 1994). Including the yellow player provided the
opportunity to create an angle that is representative with the field of view observed
by a central offensive player and at the same time indicating the position of the
particular player. No participants reported problems imagining themselves as being
the yellow player.
The simulations varied in the number of players presented: 2 vs. 1 (i.e., 2
at­tackers, including the yellow player, versus 1 defender), 3 vs. 1, 3 vs. 2, 4 vs. 3,
and 5 vs. 3. A goalkeeper was included in each condition. Figure 1 shows frames
from 3 vs. 1 and 5 vs. 3 simulations. The sequences lasted on average 6 s (range
3.0–9.7 s) and included an offensive pattern of play with variation in positions of
defenders and attackers each ending with a pass toward the player wearing the
yellow vest. The film sequences were occluded at the moment the player wearing
the yellow vest received the ball, so as to prevent participants from receiving any
feedback in relation to the decision made during the actual filming session. Five
experienced soccer coaches viewed the offensive simulations independently to
ensure that each sequence was realistic and representative of actual game play.
Only those sequences approved by all five coaches were included in the test film.
The test film included a standardized countdown from 3 to 1 preceding the pre-
sentation of each clip. Moreover, an auditory beep was presented just before the
start of each sequence.

Apparatus
Visual search data were collected using an eye-head integration system incorporating
an Applied Science Laboratories (ASL, Bedford, MA) 5000 eye movement registra-
tion system and a Flock of Birds magnetic head tracker. This is a monocular system
that measures eye line-of-gaze relative to the head. An extended-range transmitter
enabled data to be recorded up to 3 m from the magnetic head tracker providing
participants with considerable freedom of movement. The eye–head integration
system works by detecting three information characteristics: the displacement
between the pupil and corneal reflex of the left eye, the position of the eye in the
head, and the position and orientation of the head in space. The relative position
of these features enables the eye–head integration system to compute visual point
of gaze with respect to a fixed scene space. The data were collected with ASL’s
Eyepos software sampling at 60 Hz. The default criteria of ASL’s Eyenal software
were used to calculate fixation data (i.e., number of fixations and fixation ­duration).
Visual Search Behavior and Decision Making    153

Goalkeeper

Defender
Attacker
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

Yellow player
Attacker

Goalkeeper

Attacker Defender Defender

Defender Attacker

Attacker

Yellow player

Figure 1 — Top panel: Frame extracted from a 3 vs. 1 simulation. Bottom panel: Frame
extracted from a 4 vs. 3 simulation.
154   Vaeyens, Lenoir, Williams, Mazyn, and Philippaerts

A stationary scene camera was used in preference to the more commonly employed
head-mounted camera. Fixation locations were defined by comparing the point-
of-gaze cursor, superimposed on the film sequence (i.e., the integrated eye–head
data) with the coordinates obtained via the Eyenal program.

Procedure
In order to ensure that there were no baseline differences in reaction time across
groups, a standard laboratory test of choice reaction time was employed. Partici-
pants responded quickly and accurately when a stimulus light was presented after
a random foreperiod. These data were analyzed using a repeated measures ANOVA
with group as the between-participants factor and number of choices (one, two, or
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

four) as the within-participants factor.


A standardized explanation of the test procedure was then projected onto the
4.3-m × 2.5-m test screen using an LCD projector (NEC VT 670-XGA, Hong Kong).
The screen was positioned 4.4 m from the player such that a near life-size image
was presented, subtending a visual angle of around 50° in the horizontal and 30°
in the vertical direction. Participants stood on two pressure sensitive switches and
were required to make the correct tactical decision quickly and accurately when
the ball was played in the direction of the player wearing the yellow vest (i.e., the
player representing the participant on screen). In order to increase the realism of
the test setting, participants were required to respond either by passing the ball to
a player on screen, kicking the ball toward goal, or moving as if to dribble the ball
around a defender. The film sequence was occluded immediately after initiation of
each response to prevent participants from receiving feedback on task performance.
Moreover, to ensure that the intended technical skill was carried out, players were
required to verbalize their intended response immediately after each trial. The
decision times were recorded using pressure-sensitive mats located underneath
each participant’s feet and the ball, the latter being positioned at a distance of 1 m
directly in front of the participant.
After the instruction procedure, the eye movement system was fitted onto the
participant’s head. A 9-point reference grid and a stationary scene camera 5-point
grid were used to calibrate the system. The calibration procedure was checked after
the practice trials and between each of the three blocks of 11 test trials. Before each
trial, an additional rapid calibration inspection was carried out to ensure system
accuracy.
Participants received four practice trials to ensure familiarity with the experi-
mental setup. Thirty-three offensive patterns were selected for the experiment: two
2 vs. 1, ten 3 vs. 1, six 3 vs. 2, six 4 vs. 3, and nine 5 vs. 3 simulations. The order
of presentation of film clips was randomized, with a comparable proportion of
viewing conditions per block. All participants viewed clips in the same order. An
intertrial interval of approximately 45 s was employed, and the entire test session
was completed in around 45 min.

Dependent Variables and Data Analysis


The following two measures of performance were obtained.
Decision Time (DT).  This was the time from the start of the final pass toward
the yellow player to the initiation of the participant’s movement response (in milli­
Visual Search Behavior and Decision Making    155

seconds). The initiation of response was recorded when the participant raised his
foot off either of the two start pads.
Response Accuracy (RA).  Seven youth soccer coaches independently stipulated
the ideal decision (i.e., the appropriate action leading to immediate goal-scoring
opportunity) for each offensive soccer simulation. The coaches agreed on their
response on 88.9% of simulations. For each simulation, no more than two coaches
had a different coding behavior. Because these alternatives were not necessarily
incorrect—there is more than one way to score a goal—it was decided not to use
the correct–incorrect paradigm. Therefore, for each simulation, the coaches also
individually rated all remaining responses given by participants during the test,
taking into account the following guidelines:
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

• Not the most goal-oriented action, but action “indirectly” leading to a goal-
scoring opportunity = 2
• Continuation of ball possession, however, not leading to a goal scoring oppor-
tunity = 1
• A wrong decision leading to loss of ball possession = 0
Afterward, these ratings were discussed as a panel and agreement was reached
for all responses made by participants. The coaches were blind as to the identity of
the players. The ideal decision was rated a maximum score of 3. Response accuracy
was then calculated as the mean correctness of the participant’s response relative
to the prelabeled codes across all viewing conditions (i.e., 2 vs. 1, 3 vs. 1, 3 vs. 2,
4 vs. 3 or 5 vs. 3).
Three measures of visual search behavior were recorded for each trial. The
data were collected from the onset of the soccer sequence until the initiation of
the participant’s response. If a participant responded after the end of the film clip,
visual search data were only measured for the entire duration of the sequence.
Search Rate.  This measure was the mean number of visual fixations and the mean
fixation duration per second. The default criteria of ASL’s Eyenal software were
used to calculate fixation data. A fixation was defined as a condition in which the
eye remained stationary for a period equal to, or in excess of, 116.67 ms or seven
sequential gaze position samples. This value of 116.67 ms is comparable to the
fixation time criterion used by other researchers (e.g., Williams & Davids, 1998;
Savelsbergh et al., 2002). Additionally, interfixation duration (time between end of
previous fixation and start of next) and interfixation distance (in degrees of visual
angle) data were collated to provide greater insight into the scanning behaviors
employed. A measure termed interfixation rate was calculated by dividing the
interfixation distance by the interfixation duration. The interfixation rate represents
the tempo of successive fixations.
Fixation Location.  This measure was calculated as the percentage of total viewing
time spent fixating various areas of the display when trying to make a tactical deci-
sion. Video analysis of the images recorded with the stationary scene camera was
undertaken. Two researchers independently coded the data. The level of agreement
between coders was very high (>95%). Any differences in coding were clarified
through group discussion. The fixations were allocated to one of nine areas on the
screen: yellow offensive player (YP), player in possession of the ball (PB), ball (B),
156   Vaeyens, Lenoir, Williams, Mazyn, and Philippaerts

unmarked attacker (A), attacker closely marked by a defender (A/D), defender (D),
goalkeeper (K), free space (FS), and an unspecified category for fixations that did
not match with the aforementioned areas (U).
Fixation Order.  The fixation order referred to the search sequence or pattern used
by participants. The number of times per second that participants alternated their
fixations between the player in possession of the ball, some other area of the display,
and then back to the player in possession of the ball was taken as the dependent
variable (cf. Williams et al., 1994; Williams & Davids, 1998).
A mean score per viewing condition was calculated for all dependent vari-
ables. In order to facilitate comparison between groups and viewing conditions,
fixation order and search rate variables were calculated per second (cf. Martell &
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

Vickers, 2004). These variables as well as decision time and response accuracy
were analyzed separately using a repeated measures ANOVA in which group was
the between-participants variable and viewing condition (2 vs. 1, 3 vs. 1, 3 vs.
2, 4 vs. 3, 5 vs. 3) the within-participants factor. The sphericity assumption for
repeated measures ANOVA was checked using Mauchly’s test of sphericity. When
a violation of the sphericity assumption was apparent, the Greenhouse-Geisser
correction procedure was used to adjust the degrees of freedom. Significant main
effects were followed up using Tukey post hoc tests. The fixation location data
were analyzed separately for each viewing condition using a MANOVA in which
group was the fixed factor and the nine fixation locations the dependent variables.
Significant main effects were followed up using Tukey post hoc tests. The level of
significance was set at p < .05.

Results
Choice Reaction Time
No significant differences in choice reaction time were observed across groups, F(3,
80) = 0.48, p = .70. The mean reaction time values were 230.5, 229.9, 223.2, and
235.4 ms for the elite, sub-elite, regional, and nonplayer participants respectively.
Also, the Group × Number of Choices interaction was not significant, F(6, 160) =
1.30, p = .34. As expected, a significant increase in reaction time was observed as
the number of alternatives increased, F(2, 160) = 279.96, p < .001, η2 = .78.

Decision-Making Skill Test


Decision Time.  The results are presented in Table 1. Significant effects were
observed for group, F(3, 81) = 30.47, p < .001, η2 = .53, viewing condition,
F(3.37, 272.36) = 23.55, p < .001, η2 = .23, and the Group × Viewing Condition
inter­action, F(10.09, 272.36) = 6.38, p < .001, η2 = .19. Post hoc tests revealed
that the three groups of soccer players employed shorter decision times than the
nonplayer participants across all viewing conditions. The nonplayers needed almost
twice the time used by the elite and sub-elite players to make a decision. The elite
and sub-elite players significantly outperformed their regional counterparts in all
conditions, except for the 2 vs. 1 simulations.
Participants employed the fastest decision times in the 3 vs. 1 viewing condi-
tion and the slowest in the 4 vs. 3 viewing condition. These two viewing conditions
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

Table 1  Mean Decision Time (DT, in Milliseconds) and Response Accuracy (RA, %)
± SD per Group and Viewing Condition

Group

Viewing condition DT and RA   Nonplayer   Regional   Sub-Elite    Elite Group effect, F3

    2 vs. 1    DT 1,099.3 ± 405.6a 863.9 ± 272.6b 643.7 ± 242.2b 745.3 ± 188.6b    9.6***
   RA 97.5 ± 11.2 100 ± 0.0 100 ± 0.0 100 ± 0.0    n.s.

    3 vs. 1    DT 1,153.1 ± 337.8a 791.2 ± 237.9b 567.0 ± 151.8c 564.8 ± 149.7c    29.3***
a b b b
   RA 82.8 ± 14.2 94.3 ± 5.0 94.9 ± 4.2 95.4 ± 4.7    11.7***

    3 vs. 2    DT 1,337.8 ± 326.3a 884.0 ± 205.5b 690.2 ± 235.6c 647.6 ± 201.4c    33.7***
   RA 83.1 ± 12.3 83.1 ± 16.5 88.6 ± 12.6 87.3 ± 15.1    n.s.

    4 vs. 3    DT 1,397.0 ± 396.0a 977.8 ± 301.7b 757.7 ± 206.4b,c 728.9 ± 239.2c    22.5***
a b b
   RA 28.9 ± 22.5 73.2 ± 20.0 60.3 ± 31.5 70.1 ± 24.3b    13.7***

    5 vs. 3    DT 1,396.9 ± 450.6a 870.6 ± 262.8b 565.3 ± 214.0c 568.6 ± 194.2c    35.8***
a a,b b,c c
   RA 77.2 ± 13.1 82.8 ± 11.8 89.1 ± 6.5 92.6 ± 6.8    9.6***
Note. Means in the same row having the same superscript are not significantly different at p < .05.
***p < .001.
Visual Search Behavior and Decision Making    157
158   Vaeyens, Lenoir, Williams, Mazyn, and Philippaerts

were significantly different from all other simulations. The nonplayers reacted
significantly faster in the 2 vs. 1 and 3 vs. 1 viewing conditions compared to the
other viewing conditions. The decision times for the nonplayers increased as the
number of players presented on screen increased, with the exception of the 5 vs.
3 viewing condition. A similar pattern was apparent across the three groups of
soccer players; the decision time increased from 3 vs. 1 to 3 vs. 2, with the slowest
decision times being observed in the 4 vs. 3 viewing condition, before decreasing
again in the 5 vs. 3 viewing condition.
Response Accuracy.  The response accuracy scores are presented in Table 1.
There were significant effects for group, F(3, 81) = 17.74, p < .001, η2 = .40, viewing
condition, F(2.12, 171.53) = 110.80, p < .001, η2 = .58, and the Group × Viewing
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

Condition interaction, F(6.36, 171.53) = 7.48, p < .001, η2 = .22. Post hoc tests
revealed that the three groups of soccer players made more accurate decisions in
the 3 vs. 1 and 4 vs. 3 viewing conditions compared to the soccer nonplayers. Only
the elite and sub-elite players outperformed their non-soccer-playing counterparts
in the 5 vs. 3 viewing condition, whereas the elite players were significantly better
than their regional counterparts. All participants were less accurate in the 4 vs. 3
viewing condition and more accurate in the 2 vs. 1 viewing condition.
Search Rate.  ANOVA showed no significant differences between groups in the
mean number of fixations per second, F(3, 54) = 2.05, p > .05, or in the mean fixa-
tion duration, F(3, 54) = 1.36, p > .05. The Group × Viewing condition interaction
was not significant for the mean fixation duration, F(6.16, 110.85) = 1.15, p > .05,
whereas there was a trend for significance for the mean number of fixations per
second, F(9.48, 170.67) = 1.70, p = .089, η2 = .09. However, significant differences
were found in the viewing condition main effect for both the mean number of fixa-
tions per second, F(3.16, 170.67) = 24.20, p < .001, η2 = .31, and mean fixation
duration per trial, F(2.05, 110.85) = 74.51, p < .001, η2 = .58. Significant differences
were observed between the 2 vs. 1 and 3 vs. 1 viewing conditions on the one hand
and the 3 vs. 2, 4 vs. 3 and 5 vs. 3 viewing conditions on the other. The first group
of viewing conditions was characterized by a low fixation frequency combined
with a long mean fixation duration, whereas more fixations of shorter duration were
apparent in the latter group of simulations. Moreover, the mean fixation duration
in the 3 vs. 2 viewing condition was significantly higher than in the 4 vs. 3 and 5
vs. 3 viewing conditions. These data are presented in Figures 2 and 3.
ANOVA showed no significant differences in interfixation rate between groups,
F(3, 54) = 0.73, p > .05. Significant effects were observed for viewing condition,
F(2.80, 151.22) = 42.45, p < .001, η2 = .44, and the Group × Viewing condition
interaction, F(8.40, 151.22) = 2.67, p < .01, η2 = .13. When compared to the other
conditions, the tempo of successive fixations was significantly lower in the 2 vs.
1 and the 3 vs. 1 viewing conditions. Also, in the 5 vs. 3 viewing condition, the
interfixation rate was significantly slower than in the 3 vs. 2 and the 4 vs. 3 soccer
film clips. These data are presented in Figure 4.
Fixation Order.  ANOVA showed significant differences between groups in the
frequency of alteration of fixation between the player in possession of the ball and
any other area of the scene, F(3, 54) = 7.58, p < .001, η2 = .30. Post hoc tests revealed
that the elite players (0.154 ± .010) used a higher search order compared to the
other participants (nonplayer, 0.105 ± .011; regional 0.099 ± .011; sub-elite, 0.086
Visual Search Behavior and Decision Making    159
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

Figure 2 — Mean (±SE) number of fixations (per second) per group and viewing
­condition.

Figure 3 — Mean (±SE) fixation duration per group and viewing condition.
160   Vaeyens, Lenoir, Williams, Mazyn, and Philippaerts
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

Figure 4 — Mean (±SE) Interfixation rate (degrees per second) per group and viewing
condition.

± .012). Significant effects were observed for viewing condition, F(2.35, 127.12) =
10.89, p < .001, η2 = .17, and the Group × Viewing condition interaction, F(7.06,
127.12) = 2.74, p < .001, η2 = .13. The search order was significantly higher in the
3 vs. 2 viewing condition compared to the other conditions in the soccer players.
In the 3 vs. 1 viewing condition, there were significantly more alterations of fixa-
tions between player with the ball and other areas of the display than in the 5 vs. 3
viewing condition. The significant Group × Viewing condition interaction effects
are presented in Figure 5. No significant differences were apparent across the three
soccer groups, with mean values varying across viewing conditions, which is clearly
different from the relative flat pattern evident in the soccer nonplayers.
Fixation Location.  The mean (±SE) results for each viewing condition and
group are presented in Figures 6 through 10. Overall, the player in possession of
the ball received the highest percentage of fixation time. In the 2 vs. 1 and 3 vs.
1 viewing conditions participants fixated more than 80% of the time to the player
in possession of the ball, the ball, or the yellow player. When attempting to make
a decision in offensive situations involving more players (i.e., 3 vs. 2, 4 vs. 3, and
5 vs. 3), this value decreased in favor of other areas. In these situations, players
spent more time fixating other location zones (e.g., attacker, defender, and attacker
closely marked by a defender). The results of the MANOVAs are presented in
Table 2. Significant differences for the group main effect were observed across all
viewing conditions. Generally, the more skilled players spent more time fixating
the player in possession of the ball and less time on the player wearing the yellow
shirt (Figures 6–10). Post hoc tests also revealed that the elite players spent less
Visual Search Behavior and Decision Making    161
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

Figure 5 — Mean (±SE) Fixation order (times per second) per group and viewing
c­ ondition.

Figure 6 — Mean (+SE) percentage time spent viewing each fixation location for non­players
(N), regional (R), sub-elite (S) and elite (E) players in the 2 vs. 1 viewing condition. YP =
“yellow” offensive player, PB = player in possession of the ball, B = ball, A = unmarked
attacker, A/D = attacker closely marked by a defender, D = defender, K = goalkeeper, FS =
free space, U = unspecified category.
162   Vaeyens, Lenoir, Williams, Mazyn, and Philippaerts
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

Figure 7 — Mean (+SE) percentage time spent viewing each fixation location for non­players
(N), regional (R), sub-elite (S) and elite (E) players in the 3 vs. 1 viewing condition. YP =
“yellow” offensive player, PB = player in possession of the ball, B = ball, A = unmarked
attacker, A/D = attacker closely marked by a defender, D = defender, K = goalkeeper, FS =
free space, U = unspecified category.

Figure 8 — Mean (+SE) percentage time spent viewing each fixation location for non­players
(N), regional (R), sub-elite (S) and elite (E) players in the 3 vs. 2 viewing condition. YP =
“yellow” offensive player, PB = player in possession of the ball, B = ball, A = unmarked
attacker, A/D = attacker closely marked by a defender, D = defender, K = goalkeeper, FS =
free space, U = unspecified category.
Visual Search Behavior and Decision Making    163
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

Figure 9 — Mean (+SE) percentage time spent viewing each fixation location for non­players
(N), regional (R), sub-elite (S) and elite (E) players in the 4 vs. 3 viewing condition. YP =
“yellow” offensive player, PB = player in possession of the ball, B = ball, A = unmarked
attacker, A/D = attacker closely marked by a defender, D = defender, K = goalkeeper, FS =
free space, U = unspecified category.

Figure 10 — Mean (+SE) percentage time spent viewing each fixation location for non­
players (N), regional (R), sub-elite (S) and elite (E) players in the 5 vs. 3 viewing ­condition.
YP = “yellow” offensive player, PB = player in possession of the ball, B = ball, A = unmarked
attacker, A/D = attacker closely marked by a defender, D = defender, K = goalkeeper, FS =
free space, U = unspecified category.
164   Vaeyens, Lenoir, Williams, Mazyn, and Philippaerts

Table 2  MANOVA for Examining Differences in Fixation Location


Across Groups

Viewing condition Effect Λ   F    df η2


    2 vs. 1 Group .47 2.40*** 18, 139.1 .23
    3 vs. 1 Group .35 2.95*** 21, 138.4 .30
    3 vs. 2 Group .29 3.00*** 24, 136.9 .34
    4 vs. 3 Group .35 2.15** 27, 135.0 .29
    5 vs. 3 Group .29 3.01*** 24, 136.9 .34
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

**p < .005, ***p < .001.

time fixating the defensive players and unmarked attackers and more time fixating
attackers closely marked by a defender.

Discussion
In this study we examined differences in decision-making skill and visual search
behavior across different microstates of offensive play in soccer (i.e., 2 vs. 1, 3 vs.
1, 3 vs. 2, 4 vs. 3, and 5 vs. 3). There were two objectives. First, we examined how
search behavior altered as a function of the number of players presented on screen
as well as the relative proportion of offensive and defensive players. We envisaged
systematic differences in visual search behavior across the various offensive simu-
lations. In particular, based on previous research involving defensive simulations
(see Williams & Davids, 1998), we predicted that the number of players involved
in the simulation would affect measures of search rate, such as the mean fixation
duration and average number of fixations. Second, we were interested in identifying
how performers of varying skill and experience level extract relevant information
from the display in order to guide their subsequent decision making. The soccer
players were expected to outperform their nonplayer counterparts, whereas elite
players were predicted to have better decision-making skills than individuals who
performed at the sub-elite and regional levels.
As predicted, performance on the decision-making test and visual scanning
behavior varied as a specific function of the unique task constraints presented
by the different microstates of the game. In general, decision time increased and
response accuracy decreased as the number of players increased. As the number of
players involved in the simulation increased, there were more response alternatives
to consider, adding to the amount and complexity of information that needed to be
processed. A notable exception was apparent in the viewing condition involving
5 vs. 3, where faster decision times and higher levels of response accuracy were
observed compared with the 3 vs. 2 and 4 vs. 3 simulations. It appears that the
relative proportion of offensive to defensive players may equally affect decision-
making performance. In the latter viewing condition, involving 5 vs. 3, there are
two, rather than one, extra offensive players, potentially simplifying the choice of
pass option. In support of this notion, the fastest decision times were observed in the
3 vs. 1 viewing condition, where there are fewer players and the ratio of offensive
to defensive players is higher than in other simulations.
Visual Search Behavior and Decision Making    165

As reported previously using simulations of defensive play in soccer (­Williams


et al. 1994, Williams & Davids, 1998) and during “live” defensive tasks in ice hockey
(Martell & Vickers, 2004), the visual search behaviors employed varied across the
microstates of offensive play in soccer. The changes in search behavior observed
in the present study were to some extent similar to those previously reported by
researchers using a between-group design and filmed defensive simulations in
soccer (see Williams & Davids, 1998). Specifically, variation in the number of
players involved greatly affected search rate variables, such as the mean fixation
duration and number of fixations.
In the offensive simulations, a smaller number of fixations, longer mean fixa-
tion duration, and lower interfixation rate was employed in the 2 vs. 1 and 3 vs. 1
conditions compared with the other conditions. The lower search rate employed
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

in simulations involving 2 vs. 1 and 3 vs. 1 may arise because there are fewer
potentially informative areas to fixate upon and/or it may reflect a greater role for
information extraction via peripheral vision. These suggestions are reinforced by
the fact that in these conditions players spent more than 80% of the time fixating
the ball or the player in possession of the ball. Several researchers have highlighted
using a variety of methodological approaches the more effective use of “visual
pivots” by skilled performers (e.g., Williams & Davids, 1998). In time-constrained
situations, skilled players appear to fixate gaze centrally and use peripheral vision
to extract information from the positions and movements of players. Such a strategy
has several advantages. First, there is evidence to suggest that information may be
processed more quickly through peripheral vision rather than via the fovea, provid-
ing a significant advantage when under time pressure. Second, when peripheral
vision is used in this manner the need to move the eyes around the display using
saccadic eye movements is reduced—the latter being considered inactive periods
of information processing. The use of visual pivots may also be indicative of a
more context rather than target control strategy (for a discussion, see Tenenbaum
& Bar-Eli,1993).
In contrast to the 2 vs. 1 and 3 vs. 1 simulations, when attempting to make a
decision in offensive situations involving 3 vs. 2, 4 vs. 3, and 5 vs. 3 situations, a
higher number of fixations, shorter mean duration, and a shorter interfixation rate
were employed compared with other simulations. As the number of players and
potential response alternatives increased, players employed a higher search rate
in order to extract information from more disparate information sources. In these
situations, players spent more time fixating areas other than the ball or player in
possession of the ball compared with simulations involving 2 vs. 1 and 3 vs. 1.
An interesting anomaly is provided by the 5 vs. 3 simulation, where the ratio
of offensive to defensive players is slightly higher than in the 3 vs. 2 scenario.
In the 5 vs. 3 viewing condition, players employed a lower interfixation rate and
fewer alternations of fixation between the player with the ball and other areas of
the display than in the viewing condition involving 3 vs. 2, potentially indicating
a trend toward greater use of peripheral vision. This finding supports our earlier
argument that the emergent search rate is dependent on the relative proportion of
offensive and defensive players, and the strategic information conveyed by this ratio
as well as the number of players involved in each microstate. It would be interest-
ing to examine whether this trend is exaggerated in viewing conditions involving
more players (e.g., 5 vs. 4) and in simulations involving a higher ratio of offensive
to defensive players (e.g., 6 vs. 3).
166   Vaeyens, Lenoir, Williams, Mazyn, and Philippaerts

The present data potentially have important implications for coaches and those
involved in developing perceptual training programs. Because the emergent visual
behavior is dependent on the specific constraints imposed by the task, this would
suggest the need to maintain some element of specificity during practice. If different
search behaviors are evident during 2 vs. 1 and 4 vs. 3, for instance, then practice
in one context may not transfer to the other. Further research is needed to examine
issues related to specificity of practice and how this relates to the development of
perceptual-cognitive expertise in sport. It may be that the process of manipulating
constraints during practice provides a useful vehicle to facilitate the development
of more effective and adaptable search behaviors (for a more detailed discussion,
see Williams et al., 2004).
The current findings also highlight the difficulties involved when attempting to
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

enhance performance by asking learners to mirror the search patterns displayed by


experts. The manner in which performers extract information is governed by subtle
manipulations in display characteristics as well as the underlying knowledge struc-
tures that guide the search for relevant cues. The process of enhancing performance
by modeling expert performance strategies may therefore be fraught with difficulty.
Although expert performance strategies may be used to provide some useful guide-
lines for effective behavior, skilled performance only arises when performers are
provided with the opportunity to learn the most effective search behaviors through
guided-discovery or discovery learning (Smeeton et al., 2005).
In keeping with our second hypothesis, the soccer players demonstrated
superior decision-making skills compared to nonplayers across all the microstates
of offensive play. Generally, the soccer players were more accurate and faster in
making decisions than their nonplaying counterparts. The nonplayers typically made
a decision just as, or immediately after, the player wearing the yellow vest passed
the ball, whereas the soccer players decided upon the most appropriate response
before this player received the ball. These findings corroborate previous research
involving decision making in offensive situations in soccer (Helsen & Pauwels,
1993; Helsen & Starkes, 1999). As a result of extensive exposure to the domain
over many years of practice, skilled performers develop elaborate task-specific
knowledge structures, coupled with efficient encoding and retrieval processes,
which provide them with a significant advantage over less skilled players when
attempting to make appropriate decisions under time constraint (see Ericsson &
Delaney, 1999; Ericsson & Kintsch, 1995).
The elite and sub-elite players generally performed better than the regional
level players. Although the three groups of soccer players were reasonably well
matched in relation to the amount of playing experience, the more elite players
outperformed their regional counterparts. It appears that national (sub-elite) and
international (elite) level youth soccer players develop more refined and sophisti-
cated knowledge structures than regional players of comparable age and experience
level (see also Ward & Williams, 2003). The differences in decision-making skill
across groups suggest that the elite and sub-elite players are able to benefit more
than regional players from their exposure to the task domain (see also Williams
& Davids, 1995). Alternatively, it is feasible that the quality of the practice/match
play experience and the level of instruction provided by significant others, such as
coaches and mentors, may be equally responsible for the differences in decision-
making skill across groups (Ward, Hodges, Williams, & Starkes, 2004).
Visual Search Behavior and Decision Making    167

The decision-making test employed was sufficiently sensitive to discriminate


national and international level players from regional level players. In most previ-
ous studies, researchers have employed groups differentiated based on both their
playing skill (i.e., skilled vs. less skilled) and experience level (experienced vs.
inexperienced). The lack of differences between the national and international level
players may reflect the fact that such measures are not sensitive enough to dis-
criminate between players who are relatively close together on the skill continuum.
Although access to technology has improved markedly in recent years, difficulties
remain in trying to adequately capture the demands of competition under controlled
laboratory conditions (Williams & Ericsson, 2005).
It is also feasible that not all elite players are exceptional decision makers. The
skills necessary to achieve excellence in team sports such as soccer are multifaceted
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

in nature and weaknesses in one area may be compensated for by strengths in others
(Williams & Ericsson, 2005). For example, a player who is not an exceptional
good decision maker may be able to compensate by being quick and agile or by
developing excellent technical skills. This latter issue may be addressed method-
ologically in future research by selecting good and not-so-good decision makers
for more detailed investigation based on their performance on empirical measures
of decision-making skill. Although some sub-elite players may outperform their
elite counterparts on these measures, other weaknesses, for example in technical
ability, may prevent them from progressing to the higher level.
There were fewer differences than expected across groups in the visual search
behaviors employed, implying that the manner in which the information is processed
and evaluated is at least as important as the way in which it is extracted from the
display (cf. Abernethy & Russell, 1987). However, significant group effects were
observed for fixation order and fixation location. The elite players alternated their
gaze between the player in possession of the ball and any other area of the display
significantly more frequently than other participants. This difference, which nicely
illustrates the advantages of using process-tracing measures such as eye movement
recording to identify skill-based differences in the mechanisms underlying perfor-
mance, has been previously reported for 11 vs. 11 defensive simulations (Williams
et al., 1994). No significant differences were apparent across the three soccer groups
with mean values varying across conditions, which was clearly different from the
relatively flat profile displayed by the soccer nonplayers. These findings support
the argument that experts are more adept than novices in adapting their visual
search strategy according to the constraints of the task, providing some evidence to
illustrate their more efficient and pertinent search strategy (Williams et al., 2004).
In relation to fixation location, the elite players spent a higher percentage of time
fixating the player in possession of the ball and less time focusing on the player
wearing the yellow shirt.
In conclusion, in this study we examined the mechanisms underlying decision-
making skill in soccer. We analyzed differences in decision-making skill and visual
search behavior using simulations involving different microstates of the game and
groups of adolescents matched for age and playing experience. It was shown that
the visual search behavior employed during decision making varies as a function of
playing experience, skill level, and the unique constraints of the task. The number
of players involved and the relative proportion of offensive to defensive players all
play a part in constraining the emergent search behavior. We provided evidence to
168   Vaeyens, Lenoir, Williams, Mazyn, and Philippaerts

confirm that an increase in the number of players and a greater relative proportion
of offensive to defensive players changed the perceptual and cognitive demands of
the task. The measures employed were able to discriminate successfully between
soccer players and nonplayers, and more importantly, between elite and sub-elite
players and their regional counterparts. Such measures may have practical utility
for the testing and training of youth, and adult, soccer players.

Acknowledgments
The study was supported by funds from the Research Foundation–Flanders and the
Belgian National Lottery. We are grateful to all the players and coaches who participated in
the study. Thanks are also due to Davy Spiessens for technical support.
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

References
Abernethy, B. (1990). Expertise, visual search, and information pick-up in squash. Percep-
tion, 19, 63-77.
Abernethy, B., & Russell, D.G. (1987). Expert-novice differences in an applied selective
attention task. Journal of Sport Psychology, 9, 326-345.
Abernethy, B., Thomas, K.T., & Thomas, J.T. (1993). Strategies for improving understand-
ing of motor expertise (or mistakes we have made and things we have learned!!). In
J.L. Starkes & F. Allard (Eds.), Cognitive Issues in Motor Expertise (pp. 317-356).
Amsterdam: Elsevier.
De Groot, A.D. (1978). Thought and choice and chess. The Hague, the Netherlands:
Mouton.
Ericsson, K.A., & Delaney, P.F. (1999). Long-term working memory as an alternative pacity
models of working memory in everyday skilled performance. In A. Miyake & P. Shah
(Eds.), Models of working memory: Mechanisms of active maintenance and executive
control (pp. 257-297). Cambridge, UK: Cambridge University Press.
Ericsson, K.A., & Kintsch, W. (1995). Long-term working memory. Psychological Review,
102, 211-245.
French, K.E., & McPherson, S.L. (1999). Adaptations in response selection processes used
during sport competition with increasing age and expertise. International Journal of
Sport Psychology, 30, 173-179.
Harris, C.M. (1989). The ethology of saccades: A non-cognitive model. Biological cyber-
netics, 60, 401-410.
Harris, C.M., Hainline, L., Abramov, I., Lemerise, E., & Camenzuli, C. (1988). The distribu-
tion of fixation durations in infants and naive adults. Vision Research, 28, 419-432.
Helsen, W., & Pauwels, J.M. (1993). The relationship between expertise and visual informa-
tion processing in sport. In J.L. Starkes & F. Allards (Eds.), Cognitive issues in motor
expertise (pp. 109-134). Amsterdam: North-Holland.
Helsen, W.F., & Starkes, J.L. (1999). A multidimensional approach to skilled perception
and performance in sport. Applied Cognitive Psychology, 13, 1-27.
Martell, S.G., & Vickers, J.N. (2004). Gaze characteristics of elite and near-elite athletes in
ice hockey defensive tactics. Human Movement Science, 22, 689-712.
Reilly, T., Williams, A.M., Nevill, A., & Franks, A. (2000). A multidisciplinary approach to
talent identification in soccer. Journal of Sports Sciences, 18, 695-702.
Ripoll, H. (1991). The understanding-acting process in sport: The relationship between
semantic and sensorimotor processing. International Journal of Sport Psychology,
22, 221-243.
Savelsbergh, G.J.P., Williams, A.M., Van der Kamp, J., & Ward, P. (2002). Visual search, antici-
pation and expertise in soccer goalkeepers. Journal of Sports Sciences, 20, 279‑287.
Visual Search Behavior and Decision Making    169

Smeeton, N.J., Williams, A.M., Hodges, N.J., & Ward, P. (2005). The Relative Effective-
ness of Various Instructional Approaches in Developing Anticipation Skill. Journal of
Experimental Psychology: Applied, 11, 98-110.
Starkes, J.L., Helsen, W.F., & Jack, R. (2001). Expert performance in sport and dance. In
R.N. Singer, H.A. Hausenblas, & C.M. Janelle (Eds.), Handbook of Research in Sport
Psychology (pp. 174-201). New York: John Wiley.
Tenenbaum, G., & Bar-Eli, M. (1993). Decision making in sport. In R.N. Singer, M. Murphey,
and L.K. Tennant (Eds.), Handbook on Research in Sport Psychology (pp. 171-192).
New York: Macmillan.
Ward, P., Hodges, N.J., Williams, A.M., & Starkes, J.L. (2004). Deliberate practice and
expert performance: defining the path to excellence. In A.M. Williams, & N.J. Hodges
(Eds.), Skill Acquisition in Sport: Research, Theory and Practice (pp. 231-258). London:
Routledge.
Downloaded by New York University on 09/18/16, Volume 29, Article Number 2

Ward, P., & Williams, A.M. (2003). Perceptual and cognitive skill development in soccer:
The multidimensional nature of expert performance. Journal of Sport & Exercise
Psychology, 25, 93-111.
Williams, A.M. (2000). Perceptual skill in soccer: Implications for talent identification and
development. Journal of Sports Sciences, 18, 737-750.
Williams, A.M., & Burwitz, L. (1993). Advance cue utilization in soccer. In T. Reilly, J.
Clarys, & A. Stibbe, Science and Football II (pp. 239-244). London: E & FN Spon.
Williams, A.M., & Davids, K. (1995). Declarative knowledge in sport: A byproduct of
experience or a characteristic of expertise? Journal of Sport & Exercise Psychology,
17, 259-275.
Williams, A.M., & Davids, K. (1998). Visual search strategy, selective attention, and expertise
in soccer. Research Quarterly for Exercise and Sport, 69, 111-128.
Williams, A.M., Davids, K., Burwitz, L., & Williams, J.G. (1994). Visual search strategies
of experienced and inexperienced soccer players. Research Quarterly for Exercise and
Sport, 65, 127-135.
Williams, A.M., Davids, K., & Williams, J.G. (1999). Visual perception and action in sport.
London: E & FN Spon.
Williams, A.M., & Elliot, D. (1999). Anxiety and visual search strategy in karate. Journal
of Sport & Exercise Psychology, 17, 259-278.
Williams, A.M., & Ericsson, K.A. (2005). Perceptual-cognitive expertise in sport: Some
considerations when applying the expert performance approach. Human Movement
Science, 24, 283-307.
Williams, A.M., & Grant, A. (1999). Training perceptual skill in sport. International Journal
of Sport Psychology, 30, 194-220.
Williams, A.M., Hodges, N.J., North, J., & Barton, G. (2006). Perceiving patterns of play
in dynamic sport tasks: Investigating the essential information underlying skilled per-
formance. Perception, 35, 317-332.
Williams, A.M., Janelle, C.M., & Davids, K. (2004). Constraints on the search for visual
information in sport. International Journal of Sport and Exercise Psychology, 2, 301-
318.
Williams, A.M., Ward, P., Knowles, J.M., & Smeeton, N.J. (2002). Perceptual skill in a
real-world task: Training, instruction, and transfer in tennis. Journal of Experimental
Psychology: Applied, 8, 259-270.
Wright, R.D., & Ward, L.M. (1994). Shifts of visual attention: An historical and methodologi-
cal overview. Canadian Journal of Experimental Psychology, 48, 151-166.
Zelinsky, G.J., Rao, R.P.N., Hayhoe, M.M., & Ballard, D.H. (1997). Eye movements reveal
the spatiotemporal dynamics of visual search. Psychological Science, 8, 448-453.

Manuscript submitted: October 25, 2005


Revision accepted: October 13, 2006

You might also like