Video-based visual feedback to enhance motor learning in physical education—a systematic review

Motor learning is fundamental to the individual development process and is possible at all ages (Voelcker-Rehage, 2008); nevertheless, the prepubertal phase, and therefore the school period, is considered particularly suitable for introducing motor learning processes (Hirtz & Starosta, 2002). Due to this fundamental importance, the question of how movements are learned has been the subject of numerous scientific investigations and has led to various theories.

However, few studies have investigated the effectiveness of visual feedback using videos in physical education in school settings, which comprise more heterogeneous target groups in terms of performance and motivation than voluntary settings like extracurricular sports activities. Nevertheless, as effective movement learning supported by visual feedback is meaningful for physical education, related systematic guidance for teachers to provide such feedback is needed. For these reasons it is important to investigate whether visual feedback using videos is suitable for daily use in physical education classes and which methodological approach is the most effective regarding motor learning.

In their general orientation motor approaches are to be distinguished from action approaches (for an overview see Birklbauer, 2006). The approach that shaped the scientific discussion in the second half of the 20th century was the motor approach (Wewetzer, 2008). The motor approach is usually based on an information-theoretical approach. Motion representations are stored centrally and can be retrieved on demand (Schmidt, 1975). The learning process focuses on the reinforcement and parameterization of movement representations. A technical model gives orientation, and visual or verbal feedback helps to reduce errors (Wewetzer, 2008).

For the initiation of motor learning processes, feedback as information “relating to successful and incorrect aspects of movements already performed” (Olivier & Rockmann, 2003, p. 184) is an essential influencing factor in addition to observation exercises, self-directed practice, and conscious control of attention (Wulf, Shea, & Lewthwaite, 2010). Since the information is not obtained from the learner themselves, but is fed back from external sources, it is extrinsic in nature. Extrinsic information is conveyed via different motor learning tasks, which differ in terms of their objectives (modeling, movement topology or parameter learning) and the amount of information needed regarding what is to be learned (modeling > movement topology > parameter learning). Modeling focuses on spatial kinematic movement characteristics and the relative duration of partial movements. In teaching a movement topology, information regarding the movement goal (criterion value information) and how this can be achieved on the basis of the current execution (correcting information) is primarily used. Parameter learning, on the other hand, focuses on information that provides information about the current execution (actual value information) and the deviation from an ideal performance (discrepancy information) (Blischke, Marschall, Müller, & Daugs, 1999). Newell (1991) also discusses transitional information, which is information about the “what” and “how” of the movement, when the learner is aware of the current and target state, but is not able to improve his or her own performance (Blischke et al., 1999).

However, feedback unfolds its effect only in interaction with other parameters. The way the information is processed, the timing of the feedback process, and the frequency of feedback are important and partially influence each other. The amount of information to be conveyed subsequently affects the timing of feedback. In modeling, the time between trial and feedback information should not exceed 45 s (parameter learning ≤ 15 s), and the new trial should be carried out within the subsequent 120 s (parameter learning ≤ 20 s). In this context, a relative feedback frequency of 25% distributed over the learning process seems to be sufficient to implement learning processes without loss compared to a higher feedback frequency (Blischke et al., 1999; Marschall, Bund, & Wiemeyer, 2007) and to prevent guidance at the same time (Schmidt, 1991). In modeling, additional corrective information (Kernodle & Carlton, 1992), repeated observation (3–5 times), or preferably, a slow motion sequence (≤ 25% of normal speed), as well as pictures (semi-abstract line drawings) support the learning process (Blischke et al., 1999). To teach a basic movement structure, short, precise verbal cues (about 20 words) and easy-to-understand combinations of text and picture (picture above text, read from left to right) are helpful. Zetou, Kourtesis, Getsiou, Michalapoulou, and Kioumourtzoglou (2009) also point out that verbal cues must be linguistically appropriate for the age group. If only a partial movement is to be improved, it is recommended to limit the feedback process to one parameter (Blischke et al., 1999).

Although auditory, haptic, and visual feedback offer a variety of ways to provide feedback (Sigrist, Rauter, Riener, & Wolf, 2013) digitization has recently increased the focus on visual feedback methods that use digital devices with applications for video analysis, such as Coach’s Eye, the Hudl Technique, or Dartfish, to support the learning process (Kok, Komen, van Capelleveen, & van der Kamp, 2020; Korban & Künzell, 2019; Ste-Marie, Vertes, Rymal, & Martini, 2011). The combination of visual feedback methods and digital devices made complex visualization techniques practical and versatile, as well as accessible to new audiences such as physical education teachers or amateur coaches.

In their characteristics of following a technique model and trying to minimize errors through repeated externally organized practice with respect to an ideal movement, visual feedback methods following Bandura and Walters (1977) show pronounced intersections with principles of motor approaches and thus cognitive learning theories.

Visual feedback in motor learning in general has been thoroughly investigated. As a form of augmented feedback (Swinnen, 1996) visual feedback overcomes problems associated with visual perception, sensory information (auditory, tactile, or proprioceptive), and the temporal availability of task-intrinsic feedback.

Several studies indicate that visual feedback may be effective and enhance motor learning in sport settings (Clark & Ste-Marie, 2007; Rhoads, Da Matta, Larson, & Pulos, 2014), although not all research findings support this view (Emmen, Wesseling, Bootsma, Whiting, & van Wieringen, 1985; Rothstein & Arnold, 1976).

Particularly as a component of multimodal feedback, visual feedback seems to offer fruitful opportunities (Sigrist et al., 2013). With regard to the cognitive processing of information, one might use the multimodal memory model of Engelkamp (1998). According to this model, the same perceptions recorded via different sensors result in a more intensive processing and storage of information (Engelkamp, 1998; Hoffmann & Engelkamp, 2013). This does not contradict the findings of some studies that attribute an even greater potential to visual feedback than to verbal feedback (Rhoads et al., 2014). Due to their better distinctiveness, pictures are supposedly better remembered than words (Hoffmann & Engelkamp, 2013). However, opinions differ regarding the advantages of visual over verbal feedback (Kernodle, Johnson, & Arnold, 2001).

There is an ongoing discussion about the most effective method of providing visual feedback. However, the effectiveness of a feedback method as initiating, improving, and maintaining learning progress depends on many influencing factors (Rhoads et al., 2014; Sigrist et al., 2013). For instance, different visual feedback methods differ among other things in the way the information is provided (pictures or video), in the origin of the information (own or others’ presentation), as well as in the evaluation of the information (by experts or self-organized). Visual feedback can be achieved through self-modeling, by the practitioner analyzing their own movement executions, as well as through expert modeling, in which a video of an ideal movement execution is used to provide information. Sometimes a combination of both is used to superimpose the model on the movement execution (Korban & Künzell, 2019). Despite evidence that different visualization methods can be effective for self-modeling (Kelley & Miltenberger, 2016), expert modeling (Arbabi & Sarabandi, 2016), and using both methods simultaneously (Baudry, Leroy, & Chollet, 2006), reliable statements about a preferred method are difficult to make as they have rarely been compared directly.

The state of the research shows that there is evidence of the effective use of visual feedback in a number of research settings. In addition to sport settings (Guadagnoli, Holcomb, & Davis, 2002; Kelley & Miltenberger, 2016; Ste-Marie et al., 2011), positive examples can be found in school settings (Hitchcock, Dowrick, & Prater, 2003; Prater, Carter, Hitchcock, & Dowrick, 2012) or in mixed settings (Dowrick, 1999), including some with people with disabilities (Bellini & Akullian, 2007). The entirety of these findings suggests a positive potential for physical education in schools, but very few studies have addressed the specific needs and conditions of physical education; yet, a consideration of these specific needs and conditions is essential for physical education teachers in relation to making motor learning effective and suitable for everyday use.

This is also illustrated by data from the Organisation for Economic Co-operation and Development (OECD; 2020) which shows an average class size of 21 for elementary schools and 23 for secondary schools. As physical education is a compulsory subject in schools in many countries, including Germany (Ministerium für Kultus, 2016), all students, regardless of their interests or inclinations, participate in the lessons, which, in addition to limited time resources, has an impact on learners’ heterogeneity as well as their motivation. For example, many sports studies have relied on voluntary participation (Baudry et al., 2006), which is usually motivated by particular intentions that cannot be taken for granted in schools, despite motivation being an important factor in enhancing motor learning (Wulf & Lewthwaite, 2016). Schools usually do not have access to elaborate, time-consuming video editing technology (Ste-Marie et al., 2011). Furthermore, small groups (Boyer, Miltenberger, Batsche, & Fogel, 2009), consistently high performance levels (Baudry et al., 2006) and high intrinsic motivation only reflect the reality in schools to a limited extent.

As a systematic investigation of visually assisted video-based feedback in physical education has been lacking in research so far, this systematic review aims to summarize the research and answer questions regarding the general suitability of visual feedback, the best choice of methods (expert modeling and/or self-modeling), and the comparability of circumstances for daily use in physical education.

The following research questions were deemed particularly important for assessing the effectiveness of different visual video-based feedback methods, as well as the comparability of different settings:

A systematic search was carried out between November 2019 and June 2020 using ERIC, SCOPUS, and Web of Science (Web of Science Core Collection and Medline) databases by the first author. Through the Education Research Information Center (ERIC), the most recent results for the terms “feedback” and “motor learning” were determined by means of a keyword search and their content was examined for its thematic fit based on key categories. According to Table 1, the categories Task, Modality, Setting, and Population were hierarchically decisive for this first assessment. The results were ranked in terms of their relevance and the year of publication. This was followed by screening the title, abstract, and a review of the bibliography for cross-references of the most relevant studies, to identify additional keywords and synonyms according to the categories in Table 1, which formed the basis of the search strings.

The same procedure was also applied to the other databases, so that manuscripts pertinent to the research question could be identified. In a final step, the identified keywords were used as search terms (Table 2).

The search identified 11 studies in the different databases for final consideration in the review. Figure 1 shows the selection procedure. Some studies were excluded for reasons such as topic (e.g., behavior, inclusion), subject (e.g., mathematics, physics), target group (e.g., children with disabilities, preservice teachers only), or the methods used (e.g., tandem teaching, collaborative learning).

All the studies were reviewed using a three-step procedure, starting with the title and abstract and, if the content was suitable, continuing to read the full text. In the final stage, the characteristics listed in Table 3 had to be included in the studies. An additional researcher independently repeated the same process to improve the reliability of the results. The consistency in the selection of studies was 99.2%. Different views were discussed and the final decision was made on the basis of the inclusion criteria using the full text.

The following section provides an overview of the studies found using the search strings and their relevant parameters. It contains research results from 1996–2020 from seven different countries and educational systems. The learning groups were different ages (9–15 years old), had different performance backgrounds (beginner and advanced) and were faced with different types (open and closed skills) of learning tasks. There were also differences in feedback type (instant and delayed) and augmented feedback conditions (expert modeling, self-modeling, verbal feedback, and their derivatives).

Table 5 gives a more detailed overview of the extracted characteristics. The systematic literature search identified 11 studies that met all the inclusion criteria (Table 3). A quality assessment was conducted on these 11 studies (Table 4). This involved checking the studies for sample size, control group, and type of data. Only peer-reviewed articles published in English were considered. These criteria were supplemented by an assessment based on the PEDro Scale (Maher, Sherrington, Herbert, Moseley, & Elkins, 2003). Reliability was verified by another researcher (96%). Another researcher verified the assessments with a reliability of 96%. Due to the school setting, the organization of teaching in classes and the different methodological concepts, the studies had weaknesses, especially regarding blinding and concealed allocation.

The presentation of the results in Table 5 is based on the research questions. For clarity, the numbers of the studies are given instead of the author(s) name(s) in the following. Studies are numbered according their publication dates, starting with the most recent. Since the oldest study worked with different age groups (third and fifth graders) who learned different skills (an overhead pass in basketball and a forehand stroke in tennis), the results will be discussed as 11a and 11b, but still treated as a single study.

The systematic review was conducted to investigate the use of various visually augmented feedback types in physical education to determine the following: first, whether visually augmented and multimodal feedback was effective in physical education; second, whether visually augmented, multimodal feedback variants (expert modeling, self-modeling, or a combination of both) were more effective to enhance motor learning than verbally augmented feedback alone; and third, whether the conditions of the investigations were comparable to those found in regular school lessons.

This review is the first to systematically examine visual feedback methods in physical education and therefore extends existing knowledge with new findings. However, the review was limited by the small number of available research studies. In addition, in some cases missing data made a systematic evaluation difficult. There was also a lack of information regarding students under 9 years and above 15 years of age. Due to the small number of studies conducted for each age group it was difficult to draw conclusions for certain age groups. Only further findings from investigations in primary and secondary schools could allow definite conclusions to be drawn about the effectiveness of visual feedback in daily physical education. This is likely to be of particular interest, since periods before puberty may be especially beneficial for motor learning (Hirtz & Starosta, 2002), even if other perspectives exist (Solum, Lorås, & Pedersen, 2020).

In addition, less than half of the studies performed retention tests, so it was not clear to what extent improvements could be maintained. High-quality studies (e.g., Maher et al., 2003) with appropriate test procedures such as early retention tests, retention tests, and possibly also transfer tests for the development and maintenance of learning progress, as used in other studies (Krause, Buckwitz, & Olivier, 2010; Stöckel, Hartmann, & Weigelt, 2007), are required in the future to confirm the suitability of the methods discussed over time.

At this point, it should be pointed out that the data come from different educational systems, so that differences in the conception, design and purposes of physical education may have influenced the results.

Self-modeling and expert modeling that provide visual feedback in physical education seem to have the potential to improve students’ motor performance. The presented evidence suggested that visual feedback methods seem to be more effective than verbal feedback in physical education, possibly due to improved visualization, unique information, and the recognition of errors.

Heterogeneity as well as the motivational elements caused by the institutional settings of schools seem to have less influence on the learning progress, and may be enhanced in physical education compared to sports settings since the tendency to favor visual feedback is also evident in extracurricular sports.

Different influencing factors, such as the size of classes, the time available for giving feedback, the length of scheduled lessons, the available technical equipment, the digital literacy of teachers, and data protection requirements prevent reliable statements being made about the feasibility of visual feedback in daily practice. However, technical developments may help to mitigate the time-consuming nature of the feedback process and increase the practicability of implementing visual feedback in whole classes. The problem of poor digital literacy could be addressed by training qualified teachers and by taking digital literacy into account in the training of new teachers.

Future research is needed to evaluate the different methodological approaches with representative sample sizes, especially in elementary school contexts. In particular, the delivery of combined self- and expert modeling through mobile devices may be promising, especially if expert modeling is used for instruction and, later, for the superpositioning of one’s own and the target movement.