German Journal of Exercise and Sport Research

, Volume 47, Issue 2, pp 103–109

Gaze behaviour in offside decision-making in football

A field study
  • Urs Schnyder
  • Johan M. Koedijker
  • Ralf Kredel
  • Ernst-Joachim Hossner
Main Articles
  • 302 Downloads

Abstract

The current study aims to determine the relationship between gaze characteristics and decision-making of expert and near-expert assistant referees in judging offside in football. Six assistant referees with different levels of expertise judged set-played offside scenes on the football field, while their gaze behaviour was measured with a mobile eye tracker. The location of gaze, numbers of fixations and temporal characteristics of the final fixation around the decisive pass were analysed to compare the two expertise levels and response accuracies. Expert assistant referees judged more offside situations correctly than near-experts, however, without any differences in gaze patterns. Irrespective of expertise, decision accuracy was higher for trials in which the assistant referees focussed on the offside line rather than on receiving attackers or one of the other defenders at the moment of the pass. Moreover, strong trends were observed for the positive effects of an overall “quiet” gaze behaviour and, in particular, of long final fixations on correct decisions. Thus, maintaining a stable gaze on the offside line around the moment of the decisive pass should be regarded as a superior strategy for assistant referees to optimise decision-making.

Keywords

Referees Sports officiating Eye movements Soccer Error decisions 

Blickverhalten bei Abseitsentscheidungen im Fußball

Eine Feldstudie

Zusammenfassung

Ziel der aktuellen Studie ist es, die Beziehung zwischen dem Blickverhalten und der Entscheidungsfindung von hochkompetenten und weniger kompetenten Schiedsrichterassistenten bei der Abseitsentscheidung im Fußball zu bestimmen. Sechs Schiedsrichterassistenten mit unterschiedlichen Kompetenzniveaus beurteilten Standard-Abseitssituationen auf dem Fußballfeld, während ihr gesamtes Blickverhalten mittels eines mobilen Eyetrackers erfasst wurde. Die Blickposition, die Anzahl der Blickfixierungen und die zeitlichen Charakteristika der letzten Blickfixierung zur Zeit des entscheidenden Passes wurden analysiert, um die beiden Kompetenzniveaus sowie die Entscheidungsgenauigkeit zu vergleichen. Hochkompetente Schiedsrichterassistenten beurteilten mehr Abseitssituationen korrekt als weniger kompetente, jedoch ohne Unterschiede in den Blickmustern. Unabhängig von der Kompetenz war die Genauigkeit der Entscheidungen höher bei den Versuchen, in welchen sich die Schiedsrichterassistenten zur Zeit des Passes auf die Abseitslinie anstatt auf die angespielten Angreifer oder auf einen der anderen Verteidiger konzentrierten. Darüber hinaus wurde eine starke Tendenz für positive Effekte eines insgesamt „leisen“ Blickverhaltens und insbesondere einer langen letzten Blickfixierung auf die Entscheidung beobachtet. Daher sollte das Aufrechterhalten eines stabilen Blicks auf die Abseitslinie zur Zeit des entscheidenden Passes als eine überlegene Strategie für Schiedsrichterassistenten angesehen werden, um die Entscheidungsfindung zu optimieren.

Schlüsselwörter

Schiedsrichter Offiziellentätigkeit im Sport Augenbewegungen Fußball Fehlentscheidungen 

Introduction

Research has established that the most potent stressor in sports officiating is “having a bad game” or “making a bad call” (Rainey, 1995; Taylor, Daniel, Leith, & Burke, 1990). With a steady increase in the pace of ball games over the past decades (e. g. Norton, Craig, & Olds, 1999), the demands placed upon referees are higher than ever and so are the needs to understand why officiating errors occur. As it is rather surprising that referees have not received more attention in sport science thus far (Phillippe, Vallerand, Andrianarisoa, & Brunel, 2009), the present paper aims to provide basic data on assistant referees’ (AR) gaze behaviour, while making offside decisions in football.

Within the domain of sports officiating, the offside rule in football is very crucial due to its impact on the game, all the while between 8% and 26% of potential offside situations are judged incorrectly (see, amongst others, Catteeuw et al., 2010a; Catteeuw, Gilis, Wagemans, & Helsen, 2010b; Helsen, Gilis, & Weston, 2006; Oudejans, Bakker, & Beek, 2007; Oudejans et al., 2005; Oudejans et al., 2000). Despite its importance and the relatively high error rate, there is still much debate on the underlying cause ARs’ errors in this ruling. According to Law 11 of the International Football Association (FIFA), a player is in an offside position if “he is nearer to the opponents’ goal line than both the ball and the second last opponent … A player in an offside position is only penalised if, at the moment the ball is touched or is played by one of his team, he is, in the opinion of the referee, involved in active play by interfering with play, or interfering with an opponent, or gaining an advantage by being in that position” (FIFA, 2015, p. 36). Furthermore, being nearer to an opponents’ goal line indicates that “any part of his head, body or feet is nearer to his opponents’ goal line than both the ball and the second last opponent” (FIFA, 2015, p. 110).

The offside rule implies that the AR must solve two concurrent tasks, both of which may lead to an incorrect decision regarding on- or offside. First, the AR must judge the position of the involved attackers and defenders relative to each other, which could lead to a player-location detection error, that is, a spatial error. Second, the AR must determine the exact moment of passing, which could cause a temporal error, resulting in the judgement of a different constellation of players, and thus increasing the possibility of making an incorrect decision. As both spatial and temporal error sources are not mutually exclusive, both can have a marked influence on error rates of ARs. Beyond this, in potential offside situations, erroneous decision-making may be observed in two types of decision errors. The AR can raise the flag and signal offside while the attacker is actually not offside. In this case, the AR makes a flag error (FE). Alternately, the AR does not signal offside while the attacker is offside, that is, he commits a non-flag error (NFE) (Oudejans et al., 2000).

Several perceptual–cognitive explanations have been brought forth to account for errors made by ARs. One explanation is the optical error hypothesis proposed by Oudejans et al. (2000). This explanation is based on the finding that ARs are often not positioned on the offside line at the moment they need to judge the offside situation (Catteeuw et al., 2010a; Catteeuw et al., 2010b; Gilis, Helsen, Catteeuw et al., 2009; Gilis, Helsen, Catteeuw, & Wagemans, 2008; Oudejans et al., 2005). As a result, the projected image of the receiving attacker on the retina might be ahead of (or behind) the second last defender, although, in reality, he or she is actually behind (or ahead of) this player. Thus, the relative retinal positions of the involved agents do not necessarily specify the players’ absolute positions correctly and can lead to incorrect decisions. Consequently, the relative positions of the second last defender, the receiving attacker and the AR may influence the type of error committed. Empirical support for the predictions emanating from the optical error hypothesis was first provided by Oudejans et al. (2000) and later corroborated by Oudejans et al. (2005). However, although the optical error hypothesis could provide an explanation for a number of decision-making errors, it fails to account for errors in situations in which the position of the AR is perpendicular to the second last defender such that players are portrayed correctly on the retina. Thus, other mechanisms must play significant roles in causing incorrect decisions as well.

A second explanation for ARs’ errors is rooted in the inability of the visual system to locate and process all relevant visual information to regulate the offside rule correctly (Baldo, Ranvaud, & Morya, 2002; Belda Maruenda, 2004; Helsen et al., 2006; Sanabria et al., 1998). One of these limitations is demonstrated by the flash-lag hypothesis, which refers to the perceptual illusion that a moving object is perceived as leading its real position at a moment defined by a specific time marker, such as a flash (e. g. Baldo & Klein, 1995; Khurana & Nijhawan, 1995; Nijhawan, 1994, 2001). In this context, the moment the ball is passed could potentially act as a time marker that specifies the instant at which the players’ positions are judged by the AR (Baldo et al., 2002; Helsen et al., 2006). As a result, this illusion would cause involved players to be perceived as leading to their actual position at the instant defined by the moment of the pass. Under the assumption that, in general, attackers are moving forward and defenders are rather static or moving in the opposite direction, the flash-lag hypothesis would predict a systematic shift in the type of error committed towards FEs. For example, in a simulated on-field test in which ARs judged standardised set-plays, Gilis et al. (2009) showed the predicted bias towards FEs (27.6% of all onside situations) compared to NFEs (17.3% of all offside situations).

Whereas the flash-lag hypothesis postulates that systematic timing errors might occur due to limitations in the simultaneous processing of moving and static objects in the visual field, a third explanatory approach, the gaze shift hypothesis, is based on the delays caused by the need to foveate all relevant information sources. In this regard, Belda Maruenda (2004) argued that ARs would fixate on the ball such that, at the moment of the pass, the time needed to localise all the involved objects (attackers and defenders) would cause a systematic delay in judgement. In this respect, it had been already assumed by Sanabria et al. (1998) that the fixation point would not be the ball, but rather the player in possession of the ball, and that the AR would shift fixation from the receiving attacker to the second last defender at the moment of the pass to assess whether the attacker is in an offside position. Because of latencies induced by saccadic eye movements and subsequent fixations, the AR would then judge the offside situation at a later moment in time.

To date, all three hypotheses are controversial. In relation to the gaze shift hypotheses, Oudejans et al. (2000) showed with a head mounted camera that ARs do not orient their heads towards the passer at the moment of the pass, but anticipate the pass and thus shift their head orientation to the offside line in advance. However, despite the fact that movements of eye and head are correlated (Smeets, Hayhoe, & Ballard, 1996), inferences drawn from this study provide only an indirect answer to the question at hand. Similarly, in the video simulation study of Catteeuw, Helsen, Gilis, Van Roie and Wagemans (2009) in which the ARs were equipped with an eye tracker, the authors did not show fixations on the passer before, at and after the moment of the pass. However, as the ARs were seated in front of a monitor and were required to push a button to determine on- or offside, the question remains whether this gaze pattern would also be found in more natural conditions in which ARs have to make decisions on the field. Regarding this issue, Dicks, Button, and Davids (2010) demonstrated that the visual search behaviours of goal keepers responding to penalty kicks significantly differed depending on the type of stimulus (video vs. in situ) as well as on the type of response (button press vs. action). As the accumulated time the goal keepers fixated on the ball was double in the in situ compared to the video simulation conditions, this discrepancy in the relevance of the ball as the decisive stimulus displays a drastic difference between real-life and video simulation conditions.

Furthermore, the studies of Oudejans et al. (2000) and Catteeuw et al. (2009) employed rather straightforward set-played situations. Such situations could increase the possibility of adequately predicting the moment the ball is passed, potentially rendering the need for the use of saccades around the moment of the pass redundant. For more unpredictable situations, it may be the case that ARs’ gaze shifts between passer and offside line consequently contribute to errors. Thus, the question of which gaze behaviour ARs use in highly dynamic on-field situations still remains relevant.

When it comes to perceptual skills of referees in general, only a few notable studies have been reported thus far. In particular, besides the previously discussed study of Catteeuw et al. (2009), only two further studies on referees’ gaze behaviour can be found—the studies conducted by Bard, Fleury, Carrière, and Hallé (1980) on judging gymnastic routines and by Hancock and Ste-Marie (2013) on penalty decisions in ice hockey. All three studies employed an expert–novice comparison, and although all three studies reported better decision-making for the expert groups, none of the studies demonstrated significant differences in the respective visual search patterns. Thus, analysing the gaze characteristics of ARs in an on-field study would provide more information on the visual basis of expertise in decision-making of sports officials in general.

To this extent, in the present study, six ARs of different expertise levels judged set-played offside scenes in a real football stadium while their decision accuracies and gaze behaviours were recorded. Besides the allocation of the gaze vector to certain areas of interest, one of the central questions derived from the considerations above was whether gaze shifts around the moment of the pass could be the cause of errors in judging offside. As previous refereeing studies reported a lack of differences between expert and novice visual scan patterns, we did not expect significant correlations between expertise and gaze variables (Catteeuw et al., 2009). On the basis of the empirical findings reported above, more specific predictions were established for the relationship between visual search behaviour and actual performance. In particular, it was hypothesised that fixating on the offside line in advance of the moment of the pass leads to more accurate decisions than shifting gaze just before, around or after the moment of the pass. Therefore, an early final fixation onset was expected to lead to more accurate decisions. Furthermore, it was predicted that more accurate decision-making results from fewer fixations of longer durations, perhaps due to a general “quiescence” of the gaze pattern over the whole situation (Moore, Vine, Cooke, Ring, & Wilson, 2012) or to improved information processing up to the point in time at which the decision needs to be made, that is, the moment of the decisive pass (Mann, Williams, Ward, & Janelle, 2007). Finally, drawing on the rich body of evidence that the last fixation duration affects performance, a later final fixation offset and a longer final fixation duration were also expected to improve decision-making accuracy (e. g. Vickers, 2016).

Methods

Participants

Six male ARs (age: mean [M] = 32.3 years, standard deviation [SD] = 6.9 years) provided informed consent before participating in the study. Three participants (experts; age: M = 38.0 years, SD = 2.6 years) were in possession of a FIFA license and regularly officiated on the highest Swiss national level as well as on an international level. The experts had an overall experience of 11.3 years (SD = 3.1 years) with 3.0 years (SD = 2.6 years) of experience at international standard. The other three participants (near-experts; age: M = 26.7 years, SD = 3.8 years) officiated in the second or third highest Swiss Football league with an overall experience of 7.3 years (SD = 4.0 years) and for 2.3 years (SD = 2.3 years) at their qualification. The experiment was conducted in accordance with the Declaration of Helsinki.

Apparatus

The participants were fitted with a lightweight, mobile eye tracking device (EyeSeeCam; Kumar, Kohlbecher, & Schneider, 2009) that was connected to a laptop (Macbook Pro) worn in a custom backpack (Fig. 1). The EyeSeeCam uses binocular pupil detection and first order corneal reflections to calculate the point of gaze at 120 Hz. Both eye cameras as well as the scene camera (60 Hz) were mounted on a swim-goggle frame in order to minimise movements of the cameras relative to the eye whilst maximising comfort and mobility. In the EyeSeeCam system, the video recorded by the scene camera displays a cursor indicating the location of the point of gaze in the scene, which can then be used for offline analysis. In order to ensure the precision of gaze measures throughout the study, every participant was checked for recalibration after three situations. Furthermore, the scenes were filmed with a Sony DV camcorder MV4 (50 Hz) from the stands opposite the involved AR to record the actual decisions made as well as to determine the objective on- or offside constellation of the players.
Fig. 1

Participant with an eye tracking device, connected to a laptop worn in a custom backpack

Procedure

Participants judged set offside scenes performed by players from a U21 team in the highest Swiss league (FC Thun) with three attackers, three defenders, and a goalkeeper, wearing the home and away jerseys of their team, respectively. The scenes were practiced by the players before the experiment in order to present them as fluently and realistically as possible. With 12 field players at hand, players were rotated after each attack. In total, nine different attacks were played, of which each attack was performed four times leading to a total of 36 scenes. The nine attacks were constructed in accordance with the potentially relevant factors to ruling errors as derived from literature; particularly, factors related to the shift-of-gaze (short vs. long pass), the optical error (near vs. far attacker), and the flash-lag hypothesis (slow vs. fast attacker; not moving vs. moving attacker). Each test situation consisted of a certain number of passes (between two and four) between the attackers before the decisive pass to one of two potentially receiving attackers was made. The scenes differed systematically (1) in the type of pass (either a long pass, normal pass or quick one-two pass) and (2) in the behaviours of the attackers and defenders (either the receiving attackers stayed close to the defenders during the whole scene, the attackers stayed about 5 meters away from the offside line before the pass and moved towards to goal line shortly before the pass, or the attackers remained comparatively static while the defenders moved away from the goal line). The scripts for the resulting scenes had been checked for external validity in a preparatory study.

Data acquisition took place on three different days in July 2012 in order to guarantee sufficient recovery time for the players. To ensure a maximum resemblance to the real situation, the experiment was conducted in a large football stadium (Stockhorn Arena, Thun). On each date, two ARs were tested, while wearing their official refereeing outfit. The situations were presented to each participant in a randomised order. Whenever an attack resulted in an incomplete pass, it was repeated. The ARs were instructed to follow the rules as much as possible, that is, to only judge the position of the actual receiving attacker and raise their flag in order to indicate offside.

Dependent variables

Situation inclusion

Offside situations were included in the analyses if the situation led to (1) a judgeable pass, meaning a pass that can be reached by the attacker and that was not intercepted by the defence, as well as (2) suitable gaze data. These criteria resulted in full data sets for 36 scenes for Participants 1, 3, 4 and 5, for 21 scenes for Participant 2 (an expert) and for 12 scenes for Participant 6 (a near-expert). Situations were excluded for Participants 2 and 6 in which either the moment of the decisive pass was not detectable on the video footage or the gaze data was not collected successfully due to abnormal software terminations. However, due to the relatively small number of total participants, we opted to include all six participants in the analyses.

Response accuracy

The decisions made by the participants as well as the type of situation (onside or offside) were determined from the video footage captured from the stands. To this end, all scenes were evaluated by two independent raters. For the purpose of checking interrater reliability, a selection of 36 scenes was categorised with an interobserver agreement of 100% for the participant’s response and of 97% regarding the type of situation. For the small number of inconsistent ratings, the type of situation was finally determined by calling in a third rater.

Gaze behaviour

Gaze was manually analysed frame-by-fame from the start of the situation (the first pass) to the end (2 s after the moment of the decisive pass) using Kinovea software, a free and open source solution for video analysis in the sports field. Fixations were determined on the criterion that the point of gaze remained stable over a minimum of three consecutive frames (=120 ms; ≤2 ° of visual angle). For these fixations, the gaze vector was allocated to one of the following predefined areas of interest: passer, ball, second last defender (=offside line), third last defender, fourth last defender, receiving attacker, nonreceiving attacker or “other”. After raw coding, fixations on the third last and fourth last defender were collapsed into the category “defenders” and fixations on the receiving and nonreceiving attacker into the category “attackers”. These groupings are substantiated on the basis that, on the one hand, the offside line is clearly defined by the second last defender and that, on the other hand, the scripts for the scenes specified that both attackers could potentially receive the ball, whilst the remaining attacker was responsible for playing the decisive pass. Gaze variables derived from the frame-by-frame analyses, as justified above, were fixation location at the moment of the pass, numbers of fixations over the whole situation, numbers of fixations before the moment of the pass, as well as the onset, offset, and duration of the final fixation.

Statistical analyses

Due to the low number of participants, χ2 analyses were performed to compare the dependent variables with respect to expertise and decision accuracy. Further, signal detection theory (Macmillan & Creelman, 2005) was utilized for analyses as the sensitivity index, d’, provides valuable information on the ARs’ sensitivity to detect offside and the response bias, c, on the ARs’ preference to flag or to keep the flag down in case of doubt. The 95% confidence intervals were calculated to determine whether d’ and c were significantly different from zero. Finally, Wilcoxon signed-rank tests were calculated to analyse differences between correct and incorrect decisions in regards to the number of fixations in total, the number of fixations up to the moment of the decisive pass, the duration of the final fixation and the timing of the final fixation before (onset) and after (offset) the moment of the decisive pass. The phi coefficient φ and the correlation coefficient r were calculated as measures of effect size. The significance level was a priori set at 0.05 for all inferential-statistical analyses.

Results

Response accuracy

In total, the six participants took 177 decisions, 25 of which were judged incorrectly (14.1%). Of the 177 situations, the attacker was in an onside position 151 times and in an offside position 26 times. In the 151 onside situations, the ARs made 24 FEs (15.9%). Whereas in the 26 offside situations, the ARs made only 1 NFE (3.8%). Comparing decision-making accuracy between the three expert and three near-expert ARs, a χ2 test revealed that the experts made more correct decisions than the near-experts, χ2 (1, N = 177) = 4.93, p = 0.03, φ = 0.17 (see Table 1).
Table 1

Number of correct and incorrect decisions for the expert and near-expert group (N and %)

 

Group

Experts

Near-experts

Correct

85 (91.4%)

67 (79.8%)

Incorrect

8 (8.6%)

17 (20.2%)

Furthermore, experts, d’ = 3.16, 95% CI [1.91, 4.41], p = 0.91, as well as near-experts, d’ = 1.96, 95% CI [0.80, 3.11], p = 0.80, were able to discriminate between offside and onside situations, with both experts, c = 0.34, and near-experts, c = 0.17, showing a tendency to keep the flag down in doubtful situations.

Gaze location

Table 2 shows the distribution of fixation allocations to areas of interest at the moment of the decisive pass for the expert and near-expert groups as well as for correct and incorrect decisions. Of the 177 situations, the passer was only fixated 5 times and the ball only once, whereas ARs fixated on the offside line in 63.8%, the third and fourth last defender in 10.7% and one of the attackers in 22.0% of cases.
Table 2

Allocation of fixations at the moment of the decisive pass to areas of interest for the expert and near-expert groups and for correct and incorrect decisions (N and %)

 

Group

Decision

Experts

Near-experts

Correct

Incorrect

Offside line

59 (63.4%)

54 (64.3%)

101 (66.4%)

12 (48.0%)

Defenders

9 (9.7%)

10 (11.9%)

16 (10.5%)

3 (12.0%)

Attacker

20 (21.5%)

19 (22.6%)

29 (19.1%)

10 (40.0%)

Passer

5 (5.4%)

0 (0.0%)

5 (3.3%)

0 (0.0%)

Ball

0 (0.0%)

1 (1.2%)

1 (0.7%)

0 (0.0%)

Consequently, the offside line was mostly fixated at the moment of the decisive pass. To answer the question whether the offside line is also the best fixation location in terms of decision correctness, a comparison between fixation locations (offside line vs. “other than offside line”) and decision accuracy revealed a strong trend towards more accurate judgement if the offside line was fixated, χ2 (1, N = 177) = 3.17, p = 0.07, φ = 0.13. In addition, when, comparing fixation locations (offside line vs. “other than offside line”) between the two expertise groups, a χ2 test revealed that all ARs showed a similar gaze behaviour, χ2 (1, N = 177) = 0.01, p = 0.91, φ = 0.01.

Gaze dynamics

Table 3 shows the average number of fixations over the whole situation; the number of fixations before the decisive pass; onset, offset and duration of the final fixation with respect to the moment of the decisive pass, for correct and incorrect decisions as well as for experts and near-experts. Although, on a descriptive level, there appear to be trends that the ARs have a smaller total number of fixations, a smaller number of fixations up to the decisive pass, earlier stabilisation of the final fixation, longer final fixations, and later offset of the final fixation for correct than for incorrect decisions. However, respectively, Wilcoxon tests fail to reveal significant differences (fixations: Z = 1.36, p = 0.17, r = 0.56; fixations until pass: Z = 0.52, p = 0.60, r = 0.21; final fixation onset: Z = 0.31, p = 0.75, r = 0.13; final fixation offset: Z = 1.15, p = 0.25, r = 0.47; final fixation duration: Z = 1.78, p = 0.08, r = 0.73). Though considering the small number of participants, it should be noted that the resulting effect sizes were quite large and that, in particular, the test for final fixation duration just marginally missed conventional levels of significance. Thus, it seems fair to conclude that, at least by trend, decision making is positively affected by an overall more “quiet” gaze behaviour and, in particular, by a prolonged final fixation around the moment of the decisive pass.
Table 3

Number of fixations over the whole situation (N), number of fixations until the moment of the decisive pass (N); final fixation onset with respect to the moment of the decisive pass (ms; 0 = moment of the pass); final fixation offset with respect to the moment of the decisive pass (ms); and final fixation duration (ms); for the expert and near-expert group and for correct and incorrect decisions (M and SD)

 

Group

Decision

Experts

Near-experts

Correct

Incorrect

Fixations

5.28 (2.09)

5.51 (1.46)

5.35 (1.54)

5.97 (2.01)

Fixations until pass

3.38 (1.16)

3.40 (0.91)

3.39 (0.91)

3.63 (1.26)

Final fixation onset

−843 (433)

−983 (368)

−914 (342)

−849 (650)

Final fixation offset

753 (376)

732 (373)

778 (365)

574 (269)

Final fixation duration

1596 (807)

1715 (718)

1692 (666)

1423 (705)

Discussion

Proposed by Belda Maruenda (2004) and Sanabria et al. (1998), the gaze shift hypothesis states that delays due to saccadic eye movements from either the ball or the passer to the offside line after the moment of the pass cause errors in judging offside in football. The current study showed that this hypothesis cannot be empirically confirmed as ARs scarcely fixate on the passer or the ball before the moment of the decisive pass. This result may be taken as a replication of the findings reported by Catteeuw et al. (2009) for decision-making in video-simulated offside scenes; however, to the best of our knowledge, the present study is the first one to corroborate these findings in a natural setting.

With a closer look into ARs’ gaze behaviour, experts and near-experts do not seem to differ in their general strategies, as demonstrated by the lack of differences in the number of fixations and in the temporal aspects of the final fixation around the decisive pass. Although these results should be taken with consideration due to the small number of participants, it seems worthwhile to assert that these findings are perfectly in line with previous findings of Catteeuw et al. (2009), Bard et al. (1980), and Hancock and Ste-Marie (2013), who all found better decision-making for the expert groups, however, without any significant differences regarding the underlying visual search patterns.

In contrast to the analyses of expertise effects, decision accuracy ultimately depended by trend on certain gaze characteristics. In particular, ARs featured a stable gaze on the offside line around the moment of the decisive pass. This gaze strategy might be beneficial as anchoring gaze to the offside-line could allow for a more precise judgement of the attacker’s position relative to the second last defender and, simultaneously, for optimal use of peripheral vision to estimate the moment of the pass (cf., Put, Baldo, Cravo, Wagemans, & Helsen, 2013; Hüttermann, Memmert, & Simons, 2014). Furthermore, as it has been shown that auditory and visual stimuli interact (e. g. Alais & Burr, 2003), a more stable gaze could also lead to a more accurate use of the acoustic information from the foot–ball contact of the passer. However, further research seems to be needed to pinpoint the specific advantage of this gaze strategy, either for better spatial judgement of the position of the receiving attacker relative to the offside line or for better temporal judgement of the moment of the pass.

Focusing on visual stimuli other than the offside line, like on one of the attackers or the other defenders, seems to be accompanied by a decrease in decision-making accuracy. This finding suggests that some errors may have occurred due to the mislocalisation of the players involved, as the determination of the moment of the pass should not be affected by different fixation locations. The current data only allows for speculations on the mechanism underlying those spatial judgement errors. In this regard, it has been shown in computer tasks that the perceived positions of moving objects depend, amongst other factors, on the motions of all other moving stimuli in the visual field (e. g. Eagleman & Sejnowski, 2007; Whitney, 2002). Hence, such a motion-biasing model would also account for various perceptual mislocalisation phenomena, such as the flash-lag illusion. However, from an applied perspective, the crucial question would still remain whether different motion signals on the field would affect the perceived positions of attackers and defenders when judging offside on the field.

Finally, as it has been shown that ARs do not foveate on foot–ball contact at the decisive pass, the question arises of which informational basis is actually used by ARs to estimate the moment of the pass. In this context, it may be further speculated that the observing decision-makers (i. e. judges and referees) should generally be considered as a different population than the acting decision-makers (i. e. athletes). More precisely, the key to visual expertise in refereeing may not rely on the ability to fixate on task-specific, information-rich areas (like the shoulder in a tennis serve, or the standing leg and hip rotation in a penalty kick taker), but rather to apply a more global style of perceiving, not only based on foveal vision but also on other sensory cues stemming from peripheral vision or the auditory system. Thus, ARs’ ability to maintain a stable gaze on the offside line may provide the optimal basis for the integration of multisensory information.

Irrespective of speculations of underlying mechanisms, however, it can be concluded from our data that focussing on the offside line before the decisive pass is played and maintaining a stable gaze on this location generally serve as the superior strategy for ARs when it comes to making decisions in offside situations. From an applied perspective, it seems worthwhile to consider how such a gaze behaviour could be trained and to evaluate respective training protocols in order to optimise ARs’ decision accuracy in football. These questions will be empirically addressed by our research group in the near future.

Acknowledgements

The authors thank the U21 team of FC Thun under the responsibility of Rüdiger Böhm for the well-played offside scenes.

Compliance with ethical guidelines

Conflict of interest

U. Schnyder, J.M. Koedijker,R. Kredel and E.-J. Hossner declare that they have no competing interests.

This article does not contain any studies with human participants or animals performed by any of the authors.

Copyright information

© Springer-Verlag Berlin Heidelberg 2017

Authors and Affiliations

  • Urs Schnyder
    • 1
  • Johan M. Koedijker
    • 1
    • 2
  • Ralf Kredel
    • 1
  • Ernst-Joachim Hossner
    • 1
  1. 1.Institute of Sport ScienceUniversity of BernBernSwitzerland
  2. 2.Department of Human Movement Sciences, Faculty of Behavioural and Movement Sciences, MOVE Research Institute AmsterdamVrije Universiteit AmsterdamAmsterdamThe Netherlands

Personalised recommendations