An increasing number of competitive matches in professional sports are recorded on video and referee decisions are retrospectively analyzed by coaches, fans, players, and sport commentators using slow-motion replays. In March 2018, the International Football Association Board (IFAB), the game’s law-making body, approved, with immediate effect, the use of video assistant referees in association football. A video assistant referee has to check every situation to examine whether a potential clear and obvious error has been made in a match-changing situation. The video assistant relies on video clips in slow motion and/or real time for that. The video assistant referee eventually informs the main referee who then has the opportunity to review footage on the field before making a final decision. It is often acknowledged that slow motion distorts reality and can change the way body movements and intentions are perceived (Caruso et al., 2016). In the current study, we examined the impact of slow-motion video clips on the assessment of foul-play situations.
We used 60 representative video clips of tackle incidents to assess decision-making performance. The decision-making task required an assessment of the intentionality and seriousness of the foul to determine the disciplinary sanction for the offending player. As highlighted in the “Results” section, the performance of the elite referees in this study was comparable with previously reported decisional accuracy scores for referees on the field of play (Gilis, Weston, Helsen, Junge, & Dvorak, 2006; Mascarenhas, Button, O’Hare, & Dicks, 2009) and based on video replays (Gilis et al., 2006; Spitz et al., 2017). The rather low accuracy scores in this study and previous publications are probably due to the difficulty of the situations and also to the fact that the Laws of the Game (FIFA, 2016) leave room for interpretation by the referee. In fact, the difference between the concept of a careless or reckless offence or an offence with excessive force that should be considered to whistle a foul and give a yellow or red card, respectively, still remains quite vague.
Our results demonstrate that decisional accuracy of referees was not significantly different in slow motion compared to real time. An explanation for the fact that slowing down the video speed did not add to decisional accuracy might be that it reduces the fidelity and representativeness of the everyday performance context (Hettinger & Haas, 2003; Stoffregen, Pagulayan, Smart, & Bardy, 2003). Lorains et al. (2013a) put forward the same argument to explain why elite athletes made more accurate decisions under faster speed conditions: faster speeds more closely replicate the cognitive processing demands required by sport athletes who need to make crucial decisions in dynamic and time-constrained environments. The same can be said for referees and expert performers in other domains (e.g. military, police, aviation). On the other hand, for referees, it can be argued that the impact of slow motion on decisional accuracy depends on the type of decision and the type of situations that need to be assessed. For more objective assessments, such as offside decisions in case of assistant referees, slow motion might be of added value and increase decisional accuracy. Indeed, contrary to the findings with elite athletes, Gilis et al. (2009) observed that offside decisions were significantly more accurate when played at a slower compared to a faster speed. These offside decisions require a more objective assessment of spatial (i.e. how are the players positioned relative to one another) and temporal (i.e. the exact moment of the pass) landmarks. Furthermore, Spitz et al. (2017) showed that elite referees were more accurate in slow motion for technical decisions and in case of foul-play situations with multiple players involved and several potential foul plays at the same time. In these situations, slow motion might make it easier to select the relevant information, to see whether there is actual contact, and to identify the offender and exact location of the foul. However, the results of Spitz et al. also showed that slow motion does not add to decisional accuracy for typically more ambiguous tackle incidents and the determination of the disciplinary sanction for the offending player. These results were replicated in the current study and it seems that the impact of slow motion on decisional accuracy depends on the type of decision and situation.
Although highly valued, an analysis of accuracy scores (% correct decisions) may be deceptive because it is not always a good indicator of what people are doing and does not provide a complete and comprehensive account of performance differences. The decisions that yield a given accuracy level may encompass extremes of liberal and conservative biases. To gain more insight into the underlying mechanisms of slow motion on the perception of foul play and to determine how the decisions deviated from the reference decision, we performed a mixed ordinal regression analysis. As such, we were able to determine whether the speed of the video replay biased the decisions in a certain direction.
The results indicate that slow-motion video clips are associated with increased odds for choosing a higher category on the decision scale (i.e. no card, yellow card, or red card). In case of high-impact tackle incidents, there is a clear impact of slow motion, altering the judgment of the referees towards more severe disciplinary sanctions for the offending players. These results are in line with previous research investigating the way replay speed affects human judgment in the courtroom. Viewing a situation in slow motion, compared with regular speed, increased the perceived intent of a violent action (Caruso et al., 2016). A main characteristic of slow motion is that it affects the impressions of the duration over which real-time events unfold. As suggested by Caruso et al., the temporal modulation of the dynamics creates the perception that the offender has much more time to contemplate his action than he actually does. Therefore, physical contacts and violent actions might be perceived more intentionally and seriously. Indeed, we hypothesized that slow-motion replays could disrupt normal perception of causality (Michotte, 1954, 1963), which in turn could influence the perceived duration of the event (on top of the fact that it already was replayed in slow motion). In line with the reasoning outlined in Caruso et al. (2016), this would create a situation in which observers attribute more premeditation to the player’s action. More generally, this view is consistent with a framework in which humans continuously generate predictions based on the incoming sensory information, where predictions are tuned to physics and biomechanics of our world and its associated time constants (Richmond & Zacks, 2017). In the case of slow motion, these predictions are violated, which in turn might influence the subjective duration of the event. Indeed, spatiotemporal predictability has been shown to influence subjective duration such that unpredictable events are perceived to last longer. Thus, in this case, as slow motion alters the fit of the stimulus to those time constants, this could lead viewers to associate the player’s action with a higher degree of intentionality.
It should be noted that our speed manipulation might also change more central aspects such as density of information or visual saliency. The temporal manipulation of video speed might shift attention to different relevant aspects of the visual display parts because the manipulation raises perceptual saliency of different aspects (Fischer, Lowe, & Schwan, 2008). The effect we reported could thus be the result of differential focus on relevant aspects of the visual display between both speed conditions. Indeed, eye movement studies have shown that looking behavior in dynamic scene viewing shows two distinct phases. That is, an ambient style for initial explorative looking and a focal style for subsequent, more detailed scrutiny (Eisenberg & Zacks, 2016). Thus, follow-up studies are needed to further investigate the underlying mechanisms of the slow-motion effect. Not only eye-movement registration, but also verbal reports, spatial occlusion paradigms, and manipulations to equate presentation duration for both speed conditions are viable methodologies to disentangle and explain the effect of video speed on referees’ decisions. Moreover, future studies could manipulate presentation speed of the different segments within a clip (before and after the foul) independently to know whether slow motion affects the cause (pre-contact) and effect (post-contact) of the foul in a similar way.
Our findings on the biasing potential of slow-motion replays are relevant in light of current evolutions and innovations within team sports in general and association football in particular. Fans, sport commentators, the media, and (video assistant) referees more and more make use of technology and they only see videos and replays in slow motion. Despite the fact that referees on the field of play have to decide in real time, these slow-motion replays are adopted as objective representations of the foul-play situations. However, our results demonstrate that it is important to take the biasing potential that results from the artificial distortion of temporal dynamics into account when assessing foul play situations. Slow motion can make the offence look more pre-meditated than it actually was and can, for example, change the disciplinary sanction from a yellow card into a red card.
Slow motion can therefore not be seen as a valid basis of comparison and we recommend to only use real-time footage for judging the amount of risk for the opponent’s safety involved and the perceived impact/intent of a tackle. Reminding people that they are watching a slow-motion situation might not be sufficient since previous research has shown that even when people are aware that there is an incidental influencing factor (e.g. slow motion), they often do not correct sufficiently (Epley, Keysar, Van Boven, & Gilovich, 2004; Gilbert, 1989).
In this study, an expert panel of ex-referees determined the reference decisions for the situations. Both modes (real time and slow motion) were available to determine the reference decisions and we were not able to verify whether the expert panel based their reference decisions more on the slow motion rather than on the real-time presentation mode. A potential alternative interpretation of our results could thus be that the influence of slow motion on referee decisions reflects that they converge towards the expert panel decisions. We see two arguments that counter this alternative interpretation. First, it has been shown that showing an action at both regular and slow-motion speed is effective in reducing the possible biasing influence of slow motion (Caruso et al., 2016). Second, such an account would predict that the referee decisions would converge towards the reference decisions in the slow-motion condition which was arguably not the case (as depicted in Fig. 2). Future research could focus on the extent to which video speed influences the determination of the reference decisions by the expert panel. Furthermore, it would be interesting to examine the impact of several other potentially modulating variables, such as the number, the order, the duration, and the viewing angle of replays on the assessment of foul-play situations. In line with the findings of Caruso et al. (2016), we predict that the difference between slow and regular speed remains over multiple viewings of the same foul incident. On the other hand, the difference between slow motion and real time could decrease over multiple viewings of the same foul play (as referees might pick up on information after multiple viewings at regular speed that they had originally missed), but the bias could instead also get more pronounced the more often referees viewed the foul. A better insight in these aspects is definitely of interest for many people involved in modern football.
In the context of football refereeing, the distinction between perception and judgment has important implications for the nature of the errors they are making and for how to learn to correct them. For instance, with respect to offside judgment, the perceptual nature of the flag errors (unnecessarily signaling offside because the most advanced offender is perceived as being ahead of his actual position) has given rise to training programs where assistant referees are taught to cognitively compensate for their perceptual mistakes (e.g. Catteeuw, Gilis, Wagemans, & Helsen, 2010; Put, Wagemans, Spitz, Williams, & Helsen, 2015). A similar training program appears to be needed for those who watch fouls in slow motion, when it comes to making a judgment of the severity of the tackle and the required sanction (yellow versus red card).