Introduction: Wearable, head-mounted eye trackers are widely used in sport-psychology research to assess eye movements and provide insights into perceptual-cognitive changes underpinning the development of sport skill. Where possible, the assessment of eye movements should take place in the performer`s natural environment as decision making and attentional processes are constrained by the environmental cues the performer is responding to. However, quick head movements and ground impact in dynamic tasks, such as football games, affect the reliability in detecting eye features, limiting the evaluation of gaze location. We developed a new approach to deal with this issue and validated a method that uses head orientation as a proxy for gaze measurement in football. A window, using gaze coordinates, was designed to derive the gaze location in the absence of point-of-gaze references and validated against the gold standard point-of-gaze method. Methods: Forty-eight youth football players performed 6 v 6, small-sided games on a modified pitch. The raw gaze coordinates on the x- and y-axes of 6 players were plotted to analyse the dispersion of gaze. Two standard deviations from the mean were used to capture the majority of gaze behaviour (95.44%) and in turn develop an empirical coding window to classify the gaze in two gaze regions of interest, described as player-directed or ball/ground-directed. To validate the new method, footage collected from the eye tracker of 12 players were coded frame-by-frame using the empirical window and compared to the standard point-of-gaze coding. Results: The overall agreement between the two methods was 90%. However, this rate of agreement does not account for the agreement expected by chance. Cohen`s kappa value, which corrects for this, was 0.64 (p<0.001). Furthermore, the distribution of the categories in the data was skewed as 2830 belonged to one category and 597 belonged to the other. The kappa value was influenced by this prevalence issue and the adjusted-kappa (PABAK) was 0.80, indicating substantial agreement between the two methods. Discussion: Locomotor activities, in general, create noise in gaze data and this new coding method, using a window derived from gaze location coordinates, is a valid approach to overcome potential issues when assessing eye movements in dynamic tasks, such as football smallsided games. Researchers operating in the field may use this approach when dealing with poor quality data and it may encourage researchers to perform their experiments in participant`s natural environments.
© Copyright 2016 21st Annual Congress of the European College of Sport Science (ECSS), Vienna, 6. -9. July 2016. Published by University of Vienna. All rights reserved.