Chapter 15: Augmented Feedback

Think about a time when you were beginning to learn a new physical activity. How much success did you experience on your first few attempts? Most likely, you were not very successful. As you practiced, you probably had many questions that you needed someone to answer to help you better understand what you were doing wrong and what you needed to do to improve. Although you may have been able to answer many of your questions on your own as you continued to try different things while you practiced, you found that getting an answer from the instructor saved you time and energy.

This situation is an example of what we discussed in chapter 12 as typical in the early stage of learning a skill, or relearning a skill following an injury or illness. The significance of this example is that it points out that an important role played by the practitioner is to give augmented feedback to the learner to facilitate the skill acquisition process.

Consider the following situations. Suppose that you are teaching a golf swing or fitness activity to a class, helping a new student athletic trainer to tape an ankle, or working in a clinic with a patient learning to walk with an artificial limb. In each situation, the people practicing these skills can make many mistakes and will benefit from receiving augmented feedback. When they make mistakes, which they do in abundance when they are beginners, how do you know which mistakes to tell them to correct on subsequent attempts? If you had a video camera available, would you video them and then let them watch their own performances? Or would it be even more beneficial to take the videos and have them analyzed so that you could show them what their movements looked like kinematically? There are many ways to provide augmented feedback. But before you use any one of these, you should know how to implement that method most effectively and when to use it to facilitate learning.

When people perform a motor skill, they can have available to them two general types of performance-related information (i.e., feedback) that will "tell" them something about the outcome of the performance or about what caused that outcome. One is task-intrinsic feedback, which is the sensory-perceptual information that is a natural part of performing a skill. Each of our sensory systems can provide this type of feedback. We discussed three of these in chapter 6: touch, proprioception, and vision. For example, if a person throws a dart at a target on the wall, he or she receives visual task-intrinsic feedback from seeing the flight of the dart and where it lands on the target. In addition, the person receives tactile and proprioceptive task-intrinsic feedback from the head, trunk, and limbs as he or she prepares and executes the throw. Other sensory systems can also provide task-intrinsic feedback, as does the auditory system when the person hears the dart hit, or not hit, the target.

The second general type of performance-related information is in addition to task-intrinsic feedback. Although various terms have been used to identify this type of feedback (e.g., external feedback, task-extrinsic feedback), the term that will be used in this book is augmented feedback. The adjective "augmented" refers to adding to or enhancing something, which in this case involves adding to or enhancing task-intrinsic feedback. Augmented feedback enhances the task-intrinsic feedback that provides information the person's sensory system could readily detect on its own, though not necessarily reliably. For example, a teacher or coach might tell a golfer where his or her hands were positioned at the top of the swing, even though proprioceptive feedback would allow the person to feel for himself or herself where they were. In a clinical environment, a therapist might show an amputee patient EMG traces on a computer monitor to enhance the patient's own proprioceptive feedback to help the patient activate appropriate muscles when learning to operate a prosthetic device.

In other situations, augmented feedback adds information that the person cannot detect using his or her sensory system. For example, the golf teacher or coach might tell the golfer where the ball landed because the golfer was concentrating so much on keeping his or her head down during the swing that he or she did not see it after it was hit. Likewise, a therapist might tell a patient how much his or her body swayed because vestibular problems prevent the patient from being able to detect this information. In each of these situations, augmented feedback provides performance information that otherwise would not be available to the person.

Note that the term feedback is common to both categories of performance-related information described in the preceding paragraphs. As a result, it is important to consider the term "feedback" as a generic term that describes information people receive about their performance of a motor skill during or after the performance. To help conceptualize the relationship between the two general types of performance-related feedback, consider these two types of feedback as related members of the same family. Figure 15.1 graphically describes the feedback family relationships for task-intrinsic feedback and augmented feedback, as well as for the related specific types of each.

In figure 15.1, note that there are two categories of augmented feedback: knowledge of results and knowledge of performance. Each category can involve a variety of ways of presenting augmented feedback; this will be the topic of discussion later in this chapter.

Knowledge of Results (KR)

The category of augmented feedback known as knowledge of results (commonly referred to as KR) consists of externally presented information about the outcome of an attempt to perform a skill. In some situations, KR describes something about the performance outcome (i.e., result). For example, if a teacher tells a student in an archery class, "The shot was in the blue at 9 o'clock," the teacher is providing performance outcome information. Similarly, a therapist may provide a patient with a computer-generated graph indicating how close he came to achieving a target range of motion during hip internal rotation.

Sometimes, KR simply tells the performer whether he or she has achieved the performance goal. This is the case when some external device gives a "yes" or "no" signal indicating whether or not the performance goal was achieved. For example, to augment proprioceptive and visual feedback for a patient working on achieving a specific amount of knee extension, the therapist could set a buzzer to be activated when the patient achieved the goal number of degrees of movement. Although the buzzer would provide no information about how close to the goal the movement was if it had not been achieved, the patient would know that he or she had not achieved the goal unless the buzzer sounded. There are some additional examples of KR in table 15.1.

It is important to point out that we are using the term KR to refer to one type of augmented feedback. Although a person can obtain knowledge about the results of an action from his or her own sensory system, such as seeing whether the basketball missed the basket or went in, this type of performance-related information is taskintrinsic. It does not refer to the specific type of performance-related information that the term KR refers to in this text. The importance of this distinction is that it allows us to distinguish the specific influences of task-intrinsic and augmented feedback on skill learning.

Knowledge of Performance (KP)

The second category of augmented feedback is knowledge of performance (known as KP). This is information about the movement characteristics that led to the performance outcome. The important point here is that KP differs from KR in terms of which aspect of performance the information refers to. For example, in the archery situation described earlier, the teacher could provide KP by telling the student that he or she pulled the bow to the left at the release of the arrow. Here, the teacher augments the task-intrinsic feedback by telling the student what he or she did that caused the arrow to hit the target where it did.

In addition to giving KP verbally, there are various nonverbal means of providing KP. For example, video replay is a popular method of showing a person what he or she did while performing a skill. Video replay allows the person to see what he or she actually did that led to the outcome of that performance. Although video replay can show a performance outcome, it is commonly used as KP. Another means of providing KP that is increasing in popularity as computer software becomes more accessible is showing the person computer-generated kinematic characteristics of the just-completed performance. In clinical environments, therapists also use biofeedback devices to give KP. For example, a therapist can attach a buzzer to an EMG recording device so that the person hears the buzzer sound when he or she activates the appropriate muscle during the performance of an action. In each of these situations, sensory feedback is augmented in a way that informs the person about the movement characteristics associated with the outcome of an action.

Augmented feedback plays two roles in the skill learning process. One is to facilitate achievement of the action goal of the skill. Because augmented feedback provides information about the success of the skill in progress or just completed, the learner can determine whether what he or she is doing is appropriate for performing the skill correctly. Thus, the augmented feedback can help the person achieve the skill goal more quickly or more easily than he or she could without this external information.

The second role played by augmented feedback is to motivate the learner to continue striving toward a goal. In this role, the person uses augmented feedback to compare his or her own performance to a performance goal. The person then must decide to continue trying to achieve that goal, to change goals, or to stop performing the activity. This motivational role of augmented feedback is not the focus of our discussion here. Others, however, have discussed it in the motor learning literature (e.g., Little & McCullagh, 1989; Wulf, Shea, & Lewthwaite, 2010). Scholars interested in the pedagogical aspects of physical education teaching (e.g., Silverman, Woods, & Subramaniam, 1998; Solmon & Lee, 1996) increasingly are studying the effects of augmented feedback on people's motivation to engage in, or continue to engage in, physical activities. In addition, augmented feedback functions as an important factor in influencing students' perceptions of ability (e.g., Fredenberg, Lee, & Solmon, 2001). And exercise psychologists have shown that augmented feedback is influential in motivating people to adhere to exercise and rehabilitation programs (e.g., Annesi, 1998; Dishman, 1993).

When a researcher or practitioner considers the use of augmented feedback to facilitate skill acquisition, an important theoretical and practical question arises: Is augmented feedback necessary for a person to learn motor skills? The answer to this question has theoretical implications for the understanding of the nature of skill learning itself. The need, or lack of need, for augmented feedback to acquire motor skills tells us much about what characterizes the human learning system and how it functions to acquire new skills. From a practical perspective, determining the necessity for augmented feedback for skill learning can serve to guide the development and implementation of effective instructional strategies. As you will see, the answer to this question is not a simple yes or no. Instead, there are four different answers. Which one is appropriate depends on certain characteristics of the skill being learned and of the person learning the skill.

Augmented Feedback Can Be Essential for Skill Acquisition

In some skill learning situations, people, for various reasons, cannot use the task-intrinsic feedback to determine what they need to do to improve performance. As a result, augmented feedback is essential for learning. The following three types of situations describe when a person may not be able to use important task-intrinsic feedback effectively.

First, some skill performance contexts do not make critical sensory feedback available to the person. For example, when a performer cannot see a target that he or she must hit, the performer does not have important visual feedback available. In this case, augmented feedback provides critical information that is not available from the task performance environment itself.

Second, because of injury, disease, and the like, the person does not have available the sensory pathways needed to detect task-intrinsic feedback for the skill he or she is learning. For these people, augmented feedback provides this missing information.

Third, in some situations the appropriate task-intrinsic feedback provides the necessary information and the person's sensory system is capable of detecting it, but the person cannot use the feedback. For example, a person learning to extend a knee a certain distance or to throw a ball at a certain rate of speed may not be able to determine the distance moved or the rate of speed of the throw because of lack of experience. In these situations, augmented feedback can make the available task-intrinsic feedback more meaningful to the performer.

Augmented Feedback May Not Be Needed for Skill Acquisition

Some motor skills inherently provide sufficient task-intrinsic feedback, so augmented feedback is redundant. For these types of skills, learners can use their own sensory feedback systems to determine the appropriateness of their movements and make adjustments on future attempts. An experiment by Magill, Chamberlin, and Hall (1991) provides a laboratory example of this type of situation. Participants learned a coincidence-anticipation skill in which they simulated striking a moving object, which was a series of LEDs sequentially lighting along a 281 cm long trackway. As they faced the trackway, they had to use a handheld bat to knock down a small wooden barrier directly under a target LED coincident with the lighting of the target. KR was the number of msec that they contacted the barrier before or after the target lighted. Four experiments showed that participants learned this task regardless of the number of trials on which they received KR during practice. In fact, receiving KR during practice did not lead to better learning than practice without KR.

A motor skill that does not require augmented feedback to learn it has an important characteristic: a detectable external referent in the environment that the person can use to determine the appropriateness of an action. For the anticipation timing task in the Magill et al. experiment, the target and other LEDs were the external referents. The learner could see when the bat made contact with the barrier in comparison to when the target LED lighted; this enabled him or her to see the relationship between his or her own movements and the goal of those movements. It is important to note here that the learner may not be consciously aware of this relationship. The sensory system and the motor control system operate in these situations in a way that does not demand the person's conscious awareness of the environmental characteristics (see Magill, 1998). Thus, the enhancement of these characteristics by providing augmented feedback does not increase or speed up learning of the skill.

Practice condition characteristics also influence the need for augmented feedback. One of these characteristics is the existence of an observational learning situation, which we discussed in chapter 14. Two different types of observational learning situations can be influential. In one, the learner observes a skilled model perform the skill. For example, in an experiment by Magill and Schoenfelder-Zohdi (1996), people who observed a skilled demonstration learned a rhythmic gymnastics rope skill as well as those who received verbal KP after each trial. In the other situation, the learner observes other beginners practice. For example, Hebert and Landin (1994) showed that beginning tennis players who watched other beginners practice learned the tennis forehand volley as well as or better than beginning players who received verbal KP. In both of these situations, beginners were able to practice and improve without augmented feedback.

There is an interesting parallel between skill learning situations in which learners do not need augmented feedback and results of studies investigating the use of teacher feedback in physical education class settings. These studies consistently have shown low correlations for the relationship between teacher feedback and student achievement (e.g., Lee, Keh, & Magill, 1993; Silverman, Tyson, & Krampitz, 1991). This finding suggests that the amount and quality of teacher feedback is influential for improving the skills of beginners in sport skills class settings, but we should not see it as the most important variable. Other variables, such as opportunities for observational learning or the clarity of task-intrinsic feedback, appear to be capable of precluding the need for augmented feedback. Our understanding of the extent of these influences awaits further research.

Augmented Feedback Can Enhance Skill Acquisition

There are some types of motor skills that people can learn without augmented feedback, but they will learn them more quickly or perform them at a higher level if they receive augmented feedback during practice. For these skills, augmented feedback is neither essential nor redundant. Instead, it enhances the learning of these skills beyond what could be achieved without augmented feedback.

Skills in this category include those for which improvement does occur through task-intrinsic feedback alone, but because of certain skill or learner characteristics, performance improvement reaches only a certain level. One type of skill that fits this description consists of relatively simple skills for which achievement of the performance goal is initially easy to attain. An example is a movement goal of moving as quickly as possible. Initially, a person can assess if a particular attempt was faster than a previous one. However, improvement seems to stop at a certain level of performance usually because the learner's lack of experience results in his or her decreased capability to discriminate small movement-speed differences. To improve beyond this level of performance, the person requires augmented feedback.

Another type of skill for which augmented feedback enhances learning is any complex skill that requires a person to acquire an appropriate multilimb pattern of coordination. For such skills, learners can attain a certain degree of success simply by making repeated attempts to achieve the performance goal. But this goal achievement process can be speeded up with the addition of KP. More specifically, the KP that works best is information about critical components of the coordination pattern.

The best research example of this type of skill is an experiment by Wallace and Hagler (1979). Participants learned a one-hand basketball set shot with the nondominant hand, from a distance of 3.03 m from the basket, and 45° to the left side of the basket. After each shot, one group received verbal KP about errors in their stance and limb movements during the shot. Another group received only verbal encouragement after each shot. Both groups could see the outcome of each shot. Figure 15.2 depicts the results. Note that KP provided an initial boost in performance for the first fifteen trials. Then, the verbal encouragement group caught up. However, similarity in performance between the two groups lasted only about ten trials; after this point, the verbal encouragement group showed no further improvement, whereas the group receiving KP continued to improve.

Augmented Feedback Can Hinder Skill Learning

An effect of augmented feedback on skill learning that many might not expect is that it can hinder the learning process and, in some cases, actually make learning worse than it would have been otherwise. This effect is especially evident when a beginning learner becomes dependent on augmented feedback that will not be available in a test situation. Typically, the performance improvement the learner experienced during practice deteriorates in the test situation. In fact, in some situations, not only does performance deteriorate when augmented feedback is withdrawn, but the test performance is no better than if augmented feedback had not been given at all.

Dependence on augmented feedback is most likely to occur when task-intrinsic feedback is minimal or difficult to detect or interpret. In this situation, people typically substitute augmented feedback for task-intrinsic feedback because it gives them an easy-to-use guide for performing correctly.

Several types of situations can lead a person to become dependent on augmented feedback. We will discuss three later in this chapter. One involves the presentation of erroneous augmented feedback. Another situation involves the presentation of concurrent augmented feedback, which refers to giving augmented feedback while a person performs a skill. The third occurs when augmented feedback is given too frequently during practice.

In this section, we focus on important issues concerning the content of augmented feedback, and then examine several types of augmented feedback that practitioners can use. We consider five issues related to the content of augmented feedback. Each of these concerns some of the kinds of information augmented feedback may contain.

Information about Errors versus Correct Aspects of Performance

An often debated issue about augmented feedback content is whether the information should refer to the mistakes made or those aspects of the performance that are correct. Research consistently has shown that error information is more effective for facilitating skill learning, especially in terms of persistence of learning and transfer capability. This evidence supports an important hypothesis, that focusing on what is done correctly while learning a skill, especially in the early stage of learning, is not sufficient by itself to produce optimal learning. Rather, the experience the person has in correcting errors by operating on error-based augmented feedback is especially important during skill acquisition to enhance future performance of the skill in different environments and situations, as well as to enhance the capability to self-correct errors while performing the skill.

Another way of looking at this issue is to consider the different roles augmented feedback plays. Error information directs a person to change certain performance characteristics; this in turn facilitates skill acquisition. On the other hand, information indicating that the person performed certain characteristics correctly tells the person that he or she is on track in learning the skill and encourages the person to keep trying. When we consider augmented feedback from this perspective, we see that whether this feedback should be about errors or about correct aspects of performance depends on the goal of the information. Error-related information works better to facilitate skill acquisition, whereas information about correct performance serves better to motivate the person to continue. Although some researchers have argued that information about correct performance has a more direct effect on learning than has been acknowledged to date (Chiviacowsky & Wulf, 2007; Wulf et al., 2010), the primary role for information about correct performance as augmented feedback is motivational.

KR versus KP

Two relevant questions concerning the comparison of the use of KR and KP in skill learning situations are these: Do practitioners use one of these forms of augmented feedback more than the other? Do they influence skill learning in similar or different ways?

Most of the evidence addressing the first question comes from the study of physical education teachers in actual class situations. The best example is a study by Fishman and Tobey (1978). Although their study was conducted many years ago, it is representative of more recent studies, and it involves the most extensive sampling of teachers and classes of any study that has investigated this question. Fishman and Tobey observed teachers in eighty-one classes teaching a variety of physical activities. The results showed that the teachers overwhelmingly gave KP (94 percent of the time) more than KR.

An answer to the second question, concerning the relative effectiveness of KR and KP, is more difficult to provide because of the lack of sufficient and conclusive evidence from research investigating this question. The following examples of experiments provide some insight into a reasonable answer.

Two of the experiments suggest that KP is better than KR to facilitate motor skill learning. Kernodle and Carlton (1992) compared KR with videotape replays and verbally presented technique statements as KP in an experiment in which participants practiced throwing a soft, spongy ball as far as possible with the nondominant arm. KR was presented as the distance of the throw for each practice trial. The results showed that KP led to better throwing technique and distance than KR. Zubiaur, Oña, and Delgado (1999) made a similar conclusion in a study in which university students with no previous volleyball experience practiced the overhead serve in volleyball. KP was specific information about the most important error to correct as it related to action either before hitting or in hitting the ball. KR referred to the outcome of the hit in terms of the ball's spatial precision, rotation, and flight. The results indicated that KP was more influential for learning the serve.

However, a study by Silverman, Woods, and Subramaniam (1999) provided evidence for the benefit of both KR and KP in terms of how each related to how often students in physical education classes would engage in successful and unsuccessful practice trials during a class. They observed eight middle school teachers teaching two classes each in various sport-related activities. The results indicated that teacher feedback as KR and as KP showed relatively high correlations with the frequency of students engaging in successful practice trials (.64 and .67, respectively).

These studies indicate that both KR and KP can be valuable for skill learning. Consider some conditions in which each of these forms of augmented feedback would be beneficial. KR will be beneficial for skill learning for at least five reasons: (1) Learners often use KR to confirm their own assessments of the task-intrinsic feedback, even though it may be redundant with task-intrinsic feedback. (2) Learners may need KR because they cannot determine the outcome of performing a skill on the basis of the available task-intrinsic feedback. (3) Learners often use KR to motivate themselves to continue practicing the skill. (4) Providing only KR may help to establish a discovery learning practice environment in which learners are encouraged to engage in trial-and-error problem-solving activity as they acquire a skill. (5) Providing only KR may help to ensure that learners adopt an external focus of attention as they practice a skill. (See Wulf, Chiviacowsky, Schiller, & Ávila, 2010, for a discussion of the potential benefits of using augmented feedback to induce an external focus of attention during practice.)

On the other hand, KP can be especially beneficial when (1) skills must be performed according to specified movement characteristics, such as gymnastics stunts, springboard dives or ballet movements; (2) specific movement components of skills that require complex coordination must be improved or corrected; (3) the goal of the action is to produce a specific kinematic, kinetic, or muscle activity profile; (4) KR is redundant with the task-intrinsic feedback.

Qualitative versus Quantitative Information

Augmented feedback can be qualitative, quantitative, or both. If the augmented feedback involves a numerical value related to the magnitude of some performance characteristic, it is called quantitative augmented feedback. In contrast, qualitative augmented feedback is information referring to the quality of the performance characteristic without regard for the numerical values associated with it.

For verbal augmented feedback, it is easy to distinguish these types of information in performance situations. For example, a therapist helping a patient to increase gait speed could give that patient qualitative information about the latest attempt in statements such as these: "That was faster than the last time"; "That was much better"; or "You need to bend your knee more." A physical education teacher teaching a student a tennis serve could tell the student that a particular serve was "good," or "long," or could say something like this: "You made contact with the ball too far in front of you." On the other hand, the therapist could give the patient quantitative verbal augmented feedback using these words: "That time you walked 3 seconds faster than the last time," or, "You need to bend your knee 5 more degrees." The teacher could give quantitative feedback to the tennis student like this: "The serve was 6 centimeters too long," or "You made contact with the ball 10 centimeters too far in front of you."

Practitioners also can give quantitative and qualitative information in nonverbal forms of augmented feedback. For example, the therapist could give qualitative information to the patient we have described by letting him or her hear a tone when the walking speed exceeded that of the previous attempt or when the knee flexion achieved a target amount. The teacher could give the tennis student qualitative information in the form of a computer display that used a moving stick figure to show the kinematic characteristics of his or her serving motion. Those teaching motor skills often give nonverbally presented quantitative information in combination with qualitative forms. For example, the therapist could show a patient a computer-based graphic representation of his or her leg movement while walking along, displaying numerical values of the walking speeds associated with each attempt or the degree of knee flexion observed on each attempt. We could describe similar examples for the tennis student.

How do these two types of augmented feedback information influence skill learning? Although the traditional view is that quantitative augmented feedback is preferred, results from an experiment by Magill and Wood (1986) suggest a different conclusion. Each participant practiced moving his or her arm through a series of wooden barriers to produce a specific six-segment movement pattern. Each segment had its own criterion movement time, which participants had to learn. Performance for the first sixty trials showed no difference between qualitative and quantitative forms of KR. However, during the final sixty trials and on the twenty no-KR retention trials, quantitative KR resulted in better performance than qualitative.

These results suggest that people in the early stage of learning give attention primarily to the qualitative information, even when they have quantitative information available. The advantage of this attention focus is that the qualitative information provides an easier way to make a first approximation of the required movement. Put another way, this information allows learners to perform an action that is "in the ballpark" of what they need to do, which, as we discussed in chapter 12, is an important goal for the first stage of learning. After they achieve this "ballpark" capability, quantitative information becomes more valuable to them, because it enables them to refine characteristics of performing the skill that lead to more consistent and efficient achievement of the action goal. It is important to keep in mind that there are limits to learners' abilities to use quantitative feedback to change their movement patterns. Giblin, Farrow, Reid, Ball, and Abernethy (2015) recently showed that skilled junior tennis players could only execute a limited number of feedback-based instructions to improve their service action. The players' ability to implement more precise instructions was apparently limited by their kinesthetic sensitivity.

Augmented Feedback Based on Error Size

A question that has distinct practical appeal is this: How large an error should a performer make before the instructor or therapist gives augmented feedback? To many, it seems reasonable to provide feedback only when errors are large enough to warrant attention. This approach suggests that in many skill learning situations, practitioners develop performance bandwidths that establish performance error tolerance limits specifying when they will or will not give augmented feedback. When a person's performance is acceptable (i.e., within the tolerance limits of the bandwidth) the practitioner does not give feedback. But if the performance is not acceptable (i.e., the amount or type of error is outside the bandwidth) the practitioner gives feedback.

Research supports the effectiveness of the performance bandwidth approach. For example, in the first reported experiment investigating this procedure, Sherwood (1988) had participants practice a rapid elbow-flexion task with a movement-time goal of 200 msec. One group received KR about their movement-time error after every trial, regardless of the amount of error (i.e., 0 percent bandwidth). Two other groups received KR only when their error exceeded bandwidths of 5 percent and 10 percent of the goal movement time. The results of a no-KR retention test showed that the 10 percent bandwidth condition resulted in the least amount of movement time variability (i.e., variable error), whereas the 0 percent condition resulted in the most variable error. Other researchers have replicated and extended these results (e.g., Cauraugh, Chen, & Radlo, 1993; Coca-Ugrinowitsch et al., 2014; Lee, White, & Carnahan, 1990).

A practical issue concerning the use of the bandwidth technique relates to the instructions provided about the bandwidth procedure. This issue is relevant because when the learners receive no augmented feedback about their performance, the implicit message is that it was "correct." There is an instruction-related question here: Is it important that the learner explicitly be told this information, or will the learner implicitly learn this information during practice? According to the results of an experiment by Butler, Reeve, and Fischman (1996), the bandwidth technique leads to better learning when the participants know in advance that not receiving KR means they are essentially "correct."

Erroneous Augmented Feedback

One of the ways augmented feedback hinders learning is by providing people with erroneous information. While this statement may seem unnecessary because it makes such common sense, the statement gains importance when it is considered in the context of practicing a skill that can be learned without augmented feedback. In this skill learning situation, augmented feedback is redundant with the information available from task-intrinsic feedback. As a result, most people would expect that to provide augmented feedback would be a waste of time because it would not influence the learner. But research evidence shows that this is not the case, because even when augmented feedback is redundant information, beginners will use it rather than ignore it.

The first evidence of this type of effect was reported by Buekers, Magill, and Hall (1992). Participants practiced an anticipation timing task similar to the one used by Magill, Chamberlin, and Hall (1991), which was described earlier in this chapter as a task for which KR is not needed to learn the task. In the Buekers et al. experiment, three of four groups received KR after every trial. The KR was displayed on a computer monitor and indicated to the participants the direction and amount of their timing error. For one of these groups, KR was always correct. But for another group, KR was always erroneous by indicating that performance on a trial was 100 msec later than it actually was. The third KR group received correct KR for the first fifty trials, but then received the erroneous KR for the last twenty-five trials. A fourth group did not receive KR during practice.

The results (figure 15.3) showed two important findings. First, the correct-KR and the no-KR groups did not differ during the practice or the retention trials, which confirmed previous findings that augmented feedback is not needed to learn this skill. Second, the erroneous KR information led participants to learn to perform according to the KR rather than according to the task-intrinsic feedback. This latter result suggested that the participants used KR, even though it was erroneous information. Even more impressive was that the erroneous KR influenced the group that had received correct KR for fifty trials and then was switched to the erroneous KR. After the switch, this group began to perform similarly to the group that had received the incorrect KR for all the practice trials. In addition, the erroneous information not only influenced performance when it was available but also influenced retention performance one day and one week later when no KR was available. A subsequent experiment (McNevin, Magill, & Buekers, 1994) demonstrated that the erroneous KR also influenced performance on a no-KR transfer test in which participants were required to respond to a faster or slower speed than they practiced.

The preceding demonstrations of erroneous KR effects were based on laboratory tasks, but similar results have been shown for sports skills. For example, an experiment by Ford, Hodges, and Williams (2007) had skilled soccer players kicking a ball to a target, which required that the ball achieve a specific height during its flight. One group of players received erroneous KR about the height of the ball's flight during each kick by observing a prerecorded video clip of a kicked ball that reached a height that was different from their own. Results showed that the players eventually based the height of their kicks on the erroneous video feedback rather than on their own sensory feedback. Erroneous augmented feedback has also been used in clinical settings to treat people with chronic pain conditions that are thought to result from distorted representations of the painful body part. For example, presenting osteoarthritis patients with resized images of their arthritic body part has been shown to change the patients' distorted perception of the size of the body part and reduce their pain (e.g., Gilpin, Moseley, Stanton, & Newport, 2015).

Why would erroneous KR affect learning a skill for which KR is redundant information? The most likely reason appears to be that when people perform skills, they rely on augmented feedback to help them deal with their uncertainty about what the task-intrinsic feedback is telling them. For the anticipation timing and soccer kicking tasks referred to earlier, the uncertainty may exist because the visual task-intrinsic feedback is difficult to consciously observe, interpret, and use. Evidence for an uncertainty-based explanation has been demonstrated in experiments by Buekers, Magill, and Sneyers (1994), and Buekers and Magill (1995).

The important message for practitioners here is that people, especially those who are in the early stage of skill learning, will use augmented feedback when it is available, whether it is correct or not. Because of their uncertainty about how to use or interpret task-intrinsic feedback, beginners rely on augmented feedback as a critical source of information on which to base how they will make corrections on future trials. As a result, instructors need to be certain that they provide correct augmented feedback.

Most of the research on which we base our knowledge of augmented feedback and skill learning comes from laboratory experiments in which researchers gave KR to participants. Although most of the conclusions from that research also apply to KP, it is useful to look at some of the research that has investigated different types of KP.

Verbal KP

One reason practitioners give verbal KP more than verbal KR is that KP gives people more information to help them improve the movement characteristics that underlie skilled performance. One of the problems that arises with the use of verbal KP is determining the appropriate content: what to tell the person practicing the skill. This problem occurs because skills are typically complex and KP usually relates to a specific feature of skill performance. The challenge for the instructor or therapist, then, is selecting the appropriate features of the performance on which to base KP.

Selecting the skill component for KP. The first thing the practitioner must do is perform a skill analysis of the skill being practiced. This means identifying the various component parts of the skill. Then, he or she should prioritize each part in terms of how critical that part is for performing the skill. Prioritize by listing the most critical part first, then the second most critical, and so on. To determine which part is most critical, decide which part of the skill absolutely must be done properly for the person to achieve the skill's action goal. For example, for the skill of throwing a dart at a target, the most critical component is looking at the target. This part is the most critical because even if a beginner did all other parts of the skill correctly (which would be unlikely), there is a very low chance that he or she would throw the dart accurately without looking at the target. In this case, then, looking at the target would be first on the skill analysis priority list, and would be the first part of the skill assessed in determining what to give KP about. (For two examples of research reporting the use and benefit of a skill analysis approach to selecting skill components as the basis for verbal KP, see Magill & Schoenfelder-Zohdi, 1996 and Weeks & Kordus, 1998.)

Descriptive and prescriptive KP. After determining which aspect of the skill about which to give KP, the practitioner needs to decide the content of the statement to make to the learner. There are two types of verbal KP statements. A descriptive KP statement simply describes the error the performer has made. The other type, prescriptive KP, not only identifies the error, but also tells the person what to do to correct it. For example, if you tell a person, "You moved your right foot too soon," you describe only the problem. However, if you say, "You need to move your right foot at the same time as you move your right arm," you also give prescriptive information about what the person needs to do to correct the problem.

Which type of verbal KP better facilitates learning? The answer is that it depends on the stage of learning of the person practicing the skill. For example, the descriptive KP statement, "You moved your right foot too soon," would help a beginner only if he or she knew the actual time at which the right foot was supposed to move. Thus, descriptive KP statements are useful to help people improve performance only after they have learned what they need to do to make a correction. This suggests that prescriptive KP statements are more helpful for beginners. But, for the more advanced performer, a descriptive KP statement often will suffice.

Video Recordings as Augmented Feedback

The availability and use of video recordings as augmented feedback argues for the need for practitioners to know about how to use them effectively. It is common to find Internet sites and articles in professional journals that offer guidelines and suggestions for the use of video replays as feedback (e.g., Bertram, Marteniuk, & Guadagnoli, 2007; Franks & Maile, 1991; Jambor & Weekes, 1995; Trinity & Annesi, 1996). However, very little empirical research exists that establishes the effectiveness of video replays as an aid for skill acquisition. In fact, the most recent extensive review of the research literature related to the use of video recordings as a source of augmented feedback in skill learning situations was published many years ago (Rothstein & Arnold, 1976). One likely reason why more recent reviews have not been published is that the research reported since that review has supported its general conclusions rather than establish different ones.

The most significant conclusion from that review was that the critical factor for determining the effectiveness of video replays as an instructional aid was the skill level of the learner rather than the type of activity. For beginners to benefit from video replays, they require the assistance of an instructor to point out critical information. Advanced performers do not appear to need instructor aid as frequently, although discussions with skilled athletes suggest they often receive greater benefit from observing replays when they receive some form of attention-directing instructions, such as verbal cues and checklists.

A good example of research that demonstrated the benefit of having an instructor point out what the observer of the video replay should look for was reported in an experiment that compared video replay as KP to verbal KP and no KP for teaching moderately skilled golfers to hit a golf ball for distance and accuracy (Guadagnoli, Holcomb, & Davis, 2002). Three groups engaged in 90 min training sessions on each of four days: a control group, which the researchers called the "self-guided" group, practiced without any KP but could hit as many golf balls as they wanted; a verbal KP group received feedback from a PGA teaching professional throughout each session. The video KP group saw video replays of their swings throughout the sessions as well as getting the verbal KP from the teaching professional. The goal was to hit golf balls with a 7-iron as far as possible along a straight line. Performance was assessed by several distance and accuracy measures. Figure 15.4 shows the results for the "accuracy distance" measure, which was calculated by subtracting the distance a ball was off the straight target line from the total distance. This result, which is representative of the other measures, indicates the effectiveness of the video replay added to the verbal KP group for learning the golf swing: The video KP group performed better than the other two groups on a retention test given two weeks after the end of the training sessions (i.e., two-week posttest). An interesting characteristic of this study demonstrates the importance of augmented feedback in relation to the amount of practice given to the learner. Because the golfers were not limited to hitting a specified number of balls during each training session, the three groups differed in the number of balls hit during training. Surprisingly, the control group hit the most practice balls, but performed worse than the other two groups. The video KP group hit the fewest practice balls but learned the skill better than the other two groups.

Research evidence also establishes that video replays transmit certain types of performance-related information more effectively than other types. One of the best examples of an experiment that supports this conclusion was one conducted many years ago by Selder and Del Rolan (1979). They compared videotape replays and verbal augmented feedback (in the form of KP) in a study in which twelve-to-thirteen-year-old girls were learning to perform a balance beam routine. All the girls used a checklist to critically analyze their own performance after each trial. One group used verbal KP to complete the checklist; another group completed the checklist after viewing videotape replays of each trial. At the end of six weeks of practice, the videotape group scored significantly higher on the routine than the verbal KP group. More important, when each factor of the total routine score was evaluated, the videotape group scored significantly higher on only four of the eight factors: precision, execution, amplitude, and orientation and direction. The two groups did not differ on the other four: rhythm, elegance, coordination, and lightness of jumping and tumbling.

The results of this study suggest that video replay facilitates the learning of those performance features that performers can readily observe and determine how to correct on the basis of what they see on the video replay. However, for performance features that are difficult to visually discern, video replay is no more effective than verbal KP.

Movement Kinetics and Kinematics as Augmented Feedback

With the widespread availability of sophisticated technology for measuring, recording, and displaying human movement, the presentation of kinematic and kinetic information as augmented feedback has become increasingly common. Magill and Anderson (2012) have highlighted a range of different ways in which feedback on kinematic and kinetic parameters has been used to facilitate the learning of sports skills. However, the use of such feedback has been equally prevalent in clinical settings; primarily during rehabilitation from accident or injury, but increasingly as a means to prevent injuries that result from poor technique (e.g., Benjaminse et al., 2015; Crowell, Milner, Hamill, & Davis, 2010). Though we have much to learn about the kinematic and kinetic parameters on which to provide feedback and the most effective ways to present this information, recent research has shown that this type of augmented feedback can allow learners to quickly acquire patterns of coordination that were once considered very difficult, if not impossible, to learn (e.g., Kovacs, Buchanan, & Shea, 2010).

One of the first studies to investigate the use of movement kinematics did not involve a computer and was carried out many years ago. This study is important because it illustrates the historical interest in this type of feedback, it involved a real-world training situation, and it exemplifies the positive effect that kinematic information can have on skill learning. Lindahl (1945) investigated the methods used to train industrial machine operator trainees to precisely and quickly cut thin disks of tungsten with a machine that required fast, accurate, and rhythmic coordination of the hands and feet. The traditional approach to training for this job was a trial-and-error method. To assess an alternative method, Lindahl created a mechanism that would make a paper tracing of the machine operator's foot movement pattern during the cutting of each disk. During training, the trainers showed the trainees charts illustrating the correct foot action (see the top portion of figure 15.5), and periodically showed them tracings of their own foot action. The results (see the bottom portion of figure 15.5) indicated that this training method based on movement kinematic information as augmented feedback enabled the trainees to achieve production performance levels in eleven weeks, compared to the five months required by trainees who used the traditional trial-and-error method. In addition, the trainees reduced their percentage of broken cutting wheels to almost zero in twelve weeks, a level not achieved by those trained with the traditional method in less than nine months.

A comprehensive series of experiments reported by Swinnen and his colleagues serve as good examples of more recent laboratory research (Beets et al., 2012; Swinnen et al., 1990; Swinnen, Walter, Lee, & Serrien, 1993). Participants in these experiments practiced a bimanual coordination task that required them to simultaneously move two levers, but with each lever requiring a different spatial-temporal movement pattern. Kinematic information was presented as augmented feedback in the form of the angular displacement for each arm superimposed over the criterion displacements. In several experiments, the kinematic augmented feedback was compared with various other forms of augmented feedback. The results consistently demonstrated the effectiveness of the use of the displacement information. Kovacs and colleagues (e.g., Kovacs et al., 2010) have shown that these types of tasks can be learned very quickly if the augmented feedback is superimposed over a template of the to-be-learned pattern and the learner is prevented from seeing his or her arms.

An experiment by Wood, Gallagher, Martino, and Ross (1992) provides a good example of the use of graphically displayed movement kinematics for learning a sports skill. Participants practiced hitting a golf shot with a 5-iron. A commercially marketed golf computer monitored the velocity, displacement, and trajectory path of each swing as the head of the club passed over light sensors on the platform from which the ball was hit. This information was then displayed on a monitor. One group saw a template of an optimum pattern along with the kinematics; a second group did not see this template. A third group received the same kinematic information verbally in the form of numbers. A fourth group received no augmented feedback. On a retention test given one week later without augmented feedback, the group that had observed the graphic presentation of the swing kinematics along with the optimum pattern template performed best.

Another type of kinematic-related information that has been effectively used as augmented feedback is a person's center of mass (COM) or center of pressure (COP). This biomechanical type of information is most commonly used as a type of augmented feedback for people learning or improving static and dynamic balance skills. For example, a study by Taube, Leukel, and Gollhofer (2008) investigated the use of a handheld laser pointed at a target on a wall to provide augmented visual feedback of each participant's COM to assist them in maintaining upright standing balance on both rigid and unstable surfaces. Results showed that participants maintained better balance stability when they used the laser pointer device than when they did not.

A unique type of kinetic feedback was used by Chollet, Micallef, and Rabischong (1988) for swimmers. The researchers developed swimming paddles that would provide information to enable highly skilled swimmers to maintain their optimal velocity and number of arm cycles in a training session. The swimming paddles contained force sensors and sound generators that transmitted an audible signal to transmitters in a swimmer's cap. The sensors were set at a desired water-propulsion-force threshold; when the swimmer reached this threshold, the paddles produced a sound audible to the swimmer. The results showed that this device helped swimmers maintain their stroke count and swimming speed when they otherwise would have found it decreasing through the course of a long-distance practice session.

Biofeedback as Augmented Feedback

The term biofeedback refers to an augmented form of task-intrinsic feedback related to the activity of physiological processes, such as heart rate, blood pressure, muscle activity, and the like. Several forms of biofeedback have been used in motor skill learning situations. The most common is electromyographic (EMG) biofeedback, which provides information about muscle activity.

EMG biofeedback has been commonly used in physical rehabilitation settings and research. Although researchers continue to debate its effectiveness there is general agreement that it can benefit motor skill learning. (For a review and critical analysis of the use of EMG biofeedback in physical therapy see Huang, Wolf, & He, 2006.) The following two research examples illustrate some of the positive results researchers have reported for the use of EMG as a source of augmented feedback. An experiment by Brucker and Bulaeva (1996) used EMG biofeedback with long-term cervical spinal cord–injured people to determine if it would help them increase their voluntary EMG responses from the triceps during elbow extension. Results indicated that participants who experienced only one 45 min treatment session significantly increased their triceps EMG activity; and those who experienced additional treatment sessions demonstrated even further increases.

The purpose of a study by Intiso and colleagues (1994) was to determine the effectiveness of EMG biofeedback to help poststroke patients overcome foot drop of the paretic limb during the swing phase of walking. Some patients received EMG biofeedback during their physical therapy, whereas others did not. A unique characteristic of this study was the use of gait analysis to assess foot drop during the gait cycle. Results of this analysis demonstrated that the EMG biofeedback intervention led to better recovery than physical therapy without the biofeedback.

Finally, another type of biofeedback has been applied in the training of competitive rifle shooters (Daniels & Landers, 1981). Heartbeat biofeedback was presented audibly to help these athletes learn to squeeze the rifle trigger between heartbeats, which is a characteristic of elite shooters.

In general, research evidence has supported the effectiveness of biofeedback as a means of facilitating motor skill learning. However, debate continues concerning the specific situations in which the use of biofeedback is an effective and preferred form of augmented feedback, especially in physical rehabilitation situations (e.g., Moreland & Thomson, 1994; Moreland, Thomson, & Fuoco, 1998).

An additional concern about the use of biofeedback as augmented feedback is the tendency for people to become dependent on it to help them perform the skill. The result of this dependency is that when they must perform the skill without biofeedback, their performance level decreases. This concern is not unique to the use of biofeedback but to all types of augmented feedback. We will address this issue, along with strategies to overcome the problem, in the last section of this chapter in which the frequency of presenting augmented feedback will be discussed.

Finally, three important questions arise about the timing of giving augmented feedback. To continue with the example in which you are teaching a person to play golf, one question is this: Should you give augmented feedback while the person swings, after he or she has hit the ball, or both times? If you give feedback after the person hits the ball, the second question arises: How soon after the person's hitting of the ball should you give augmented feedback? The third question concerns whether you should give augmented feedback every time the person hits the ball, or only a few times during the practice session.

Concurrent and Terminal Augmented Feedback

Is it better to give augmented feedback while a person is performing a skill, in what is known as concurrent augmented feedback or at the end of a practice attempt, in what we call terminal augmented feedback? Unfortunately, a search through the motor learning research literature suggests that there is no unequivocal answer to this question. However, a guideline emerges from that literature that can help us answer the question. Terminal augmented feedback can be effective in almost any skill learning situation, although the teacher or therapist must consider the nature of its effect in light of our discussion earlier in this chapter of the four different effects augmented feedback can have on skill learning. Concurrent augmented feedback, as you will see in the following sections of this discussion, seems to be most effective when the task-intrinsic feedback is difficult to use to determine how to perform the skill or improve performance. Because most of the discussion thus far in this chapter has involved terminal augmented feedback, the focus of this section will be on concurrent augmented feedback.

Effects of concurrent augmented feedback on learning. Research has shown two types of effects for the use of concurrent augmented feedback in skill learning situations. The more common is a negative learning effect. Although performance improves very well during practice when the feedback is available, it declines on retention or transfer trials during which the augmented feedback is removed. In these situations, the concurrent augmented feedback influences learners to direct their attention away from the critical task-intrinsic feedback and toward the augmented feedback. The result is that they substitute the information derived from augmented feedback for the important information they need to acquire from task-intrinsic feedback. The result is that the augmented feedback becomes an integral part of what is learned, and therefore necessary for future performance which means when it is not available, performance is worse than when it is available.

Two experiments provide examples of research that has demonstrated this negative learning effect. Verschueren, Swinnen, Dom, and DeWeerdt (1997) had elderly healthy adults and Parkinson's patients practice the continuous bimanual coordination task described in figure 11.4 in chapter 11. The task required them to learn to move two levers simultaneously for 20 sec in such a way that they would draw ellipses on the computer monitor. During the practice trials, participants saw on the monitor the drawings produced as they moved their arms. The results showed that participants in both groups made considerable improvement during practice when the concurrent visual augmented feedback was available. But their performance dropped dramatically on retention test trials without the augmented feedback.

The negative learning effect has also been demonstrated for the learning of a discrete task. Vander Linden, Cauraugh, and Greene (1993) compared concurrent and terminal augmented feedback for learning a 5 sec isometric elbow-extension force production task, which is illustrated in figure 15.6. Participants received augmented feedback by seeing the force produced during the performance of the task on each trial. One group received this feedback concurrently. Two other groups received this feedback after they completed performing the task. One of these two groups saw this information after every trial; the other saw it after every other trial. During the practice trials, the concurrent augmented feedback group performed the task better than the two terminal groups. However, 48 hr later on a retention test with no augmented feedback, the concurrent group's performance declined to a level that was the poorest of the three groups.

The second general effect is that concurrent augmented feedback enhances skill learning. A variety of situations have produced this effect. Some of these are the training of flight skills for airplane pilots (e.g., Lintern, 1991), the rehabilitation of motor skills in physical therapy (e.g., Intiso et al., 1994), the activation of specific muscles or muscle groups (e.g., Brucker & Bulaeva, 1996), and the learning of certain types of bimanual coordination laboratory tasks (e.g., Swinnen et al., 1993). In these experiments, concurrent augmented feedback enhanced relevant features of the task-intrinsic feedback that were difficult to discern without the enhancement. Because most of these experiments are described elsewhere in this book, there is no need to describe them here.

Predicting learning effects of concurrent augmented feedback. Two related hypotheses have been proposed to help us better understand how to predict when concurrent augmented feedback will have a positive or negative effect on learning. First, Annett (1959, 1969, 1970) stated that augmented feedback should be considered in terms of its information value, which he related to the "informativeness" of the task-intrinsic feedback and the augmented feedback. When the information value of task-intrinsic feedback is low, but the information value of the augmented feedback is high, a dependency on the augmented feedback is likely to develop.

Lintern and his colleagues (Lintern, 1991; Lintern, Roscoe, & Sivier, 1990) added another dimension to Annett's hypothesis by proposing that practicing with augmented feedback will benefit learning to the extent that the feedback sensitizes the learner to properties or relationships in the task that specify how the system being learned can be controlled. This means that for concurrent augmented feedback to be effective, it must facilitate the learning of the critical characteristics or relationships in the task as specified by the task-intrinsic feedback. Negative learning effects will result when the augmented feedback distracts attention away from these features. But positive learning effects will result when the augmented feedback directs attention to these features.

A recent experiment by Buchanan and Wang (2012) shows that very subtle variations in the way that concurrent augmented feedback is provided can dramatically reduce the likelihood that learners will become dependent on the feedback. Participants were required to learn a complex pattern of coordination between the two hands with the aid of a visual display that provided a Lissajous template of the required pattern and a cursor representing the relative position of the two hands. One group (Behind) practiced the task with the cursor superimposed on the template and the other group (Side) practiced the task with the cursor presented in a separate window adjacent to the template. In two experiments, both groups acquired the new pattern of coordination quickly during a 5 min practice session. However, on a retention test given 15 minutes later, the Behind group showed a dramatic drop in performance when the visual feedback was removed, whereas the Side group was able to maintain their level of performance in the absence of augmented feedback. These results show that the strong guidance-like properties of concurrent visual feedback can be exploited to facilitate performance and learning if the feedback is displayed in the appropriate manner.

The second timing issue related to augmented feedback concerns when the feedback is given terminally. Two intervals of time are created between two trials: the KR-delay interval and the post-KR interval.¹ These intervals are depicted graphically in figure 15.7. To understand the relationship between these intervals and skill learning, we must understand the influence of two variables: time, or the length of the interval, and activity, or the cognitive and/or motor activity during the interval.

The Length of the KR-Delay Interval

It is not uncommon to see statements in textbooks indicating that a learner should receive augmented feedback as soon as possible after performing a skill, because delaying it beyond a certain amount of time would lead to poorer learning. A significant problem with this viewpoint is that it has little research evidence to support it. Such a view comes from research based predominantly on animal learning (see Adams, 1987). Research has established that humans use augmented feedback as more than a reward: augmented feedback has informational value that humans use to solve problems associated with learning a skill. Whereas animal learning studies have shown that delaying reward leads to decreased learning, human skill learning studies have shown that delaying augmented feedback does not have this negative effect.

Activity during the KR-Delay Interval

Researchers investigating the effects of activity during the KR-delay interval have found three types of outcomes. The most common effect of activity during the KR-delay interval on skill learning is that it has no influence on learning. Experiments have demonstrated this result since the 1960s (e.g., Bilodeau, 1969; Boulter, 1964; Marteniuk, 1986).

The second type of effect is much less common. There is some evidence, although it is sparse, that activity during the KR-delay interval hinders learning. Two specific types of activities have shown this negative effect. One type includes activities that involve the same learning processes required by the primary task being learned. For example, Marteniuk (1986) showed that when another motor or cognitive skill had to be learned during the KR-delay interval, these activities interfered with learning of the primary skill. The other type of activity that research has shown to hinder skill learning involved estimating the movement-time error of another person's movement, which the second person performed during the interval. In an experiment by Swinnen (1990), people learned to move a lever a specified distance, involving two reversals of direction, in a criterion movement time. Participants who engaged in the error estimation activity during the KR-delay interval showed worse performance on a retention test than those who did nothing or who performed a nonlearning task during the interval.

Subjective performance evaluation. The third type of effect is that certain activities during the KR-delay interval actually can benefit learning. One type of activity that has consistently demonstrated this effect requires the person to evaluate his or her own performance. We will refer to this activity as the subjective performance evaluation strategy. Research has established the effectiveness of two approaches to the use of this strategy. One requires the estimation of the outcome of the performance, the other requires the estimation of the movement-related characteristics of the performance of the skill. Swinnen (1990), in the experiment described above, compared the strategies of the participant estimating his or her own performance outcome error with estimating performance outcome error of another person's movement. Figure 15.8 shows that subjective performance estimation led to a learning benefit, but estimating another person's performance hindered learning. More recently, Sherwood (2008) further established the learning benefit of the subjective performance error strategy in an experiment involving four different laboratory learning tasks, in which participants estimated their own error on each trial prior to receiving KR.

The second type of beneficial subjective performance estimation involves the estimation of specific characteristics of some of the movement-related components of an action. In an experiment by Liu and Wrisberg (1997) participants practiced throwing a ball at a target as accurately as possible with the nonpreferred arm and without vision of the target. Participants received KR by seeing where on the target the ball had landed on each trial. During the KR-delay interval, some of the participants rated on 5-point scales the appropriateness of the force, angle of ball release, and ball trajectory of the throw, and then estimated the throw's point value on the target. The results indicated that these participants performed more accurately on a retention test than those who had not used the estimation strategy.

We find an interesting parallel to the subjective performance evaluation strategy in what researchers call the trials-delay procedure. Here, learners receive KR for a trial after they complete performance on a later trial. Anderson, Magill, and Sekiya (1994) provided evidence for the effectiveness of this procedure. Participants practiced making a blindfolded aiming movement. One group received KR about distance error after every trial (delay-0). A second group received KR two trials later (delay-2), which meant that they were told their error for trial 1 after completing trial 3. Results (figure 15.9) were that while the delay condition hindered performance during practice, it led to better performance on a 24 hr retention test.

Learning processes. What do these different effects of activity reveal about learning processes that occur during the KR-delay interval? During this time interval, the learner is actively engaged in learning processes involving activities such as developing an understanding of the task-intrinsic feedback and establishing essential error detection capabilities.

For instructional purposes, the most significant implication of the effects of activity during the KR-delay interval is that people who are practicing a skill can use a beneficial strategy after they complete the performance of a skill and before they receive augmented feedback. That is, they could verbally describe what they think they did wrong that led to a less than desired performance outcome.

The Length of the Post-KR Interval

The significance of the post-KR interval is that it is during this period of time that the learner develops a plan of action for the next trial. This planning occurs during this interval because the learner now has available both task-intrinsic feedback and augmented feedback.

If the learner processes critical skill learning information during the post-KR interval, we would expect there to be a minimum length for this interval. Although there is not an abundance of research that has investigated this issue, there is empirical evidence from many years ago that provided evidence that indeed, this interval can be too short (e.g., Gallagher & Thomas, 1980; Rogers, 1974; Weinberg, Guy, & Tupper, 1964). For optimal learning, a minimum amount of time is needed for the learner to engage in the learning processes required. Conversely, there is no evidence indicating an optimal length for the post-KR interval. Research consistently has shown no apparent upper limit for the length of this interval.

Activity during the Post-KR Interval

The effect of engaging in activity is similar for the post-KR interval to that for the KR-delay interval. Depending on the type of activity, activity can have no effect on learning, interfere with learning, or benefit learning.

The most common finding has been that activity during the post-KR interval has no effect on skill learning. The best example is in an experiment by Lee and Magill (1983) in which participants practiced making an arm movement in 1,050 msec. During each post-KR interval, one group attempted the same movement in 1,350 msec, one group engaged in a cognitive activity involving number guessing, and a third group did no activity. At the end of the practice trials, the two activity groups showed poorer performance than the no-activity group. However, this was a temporary performance effect rather than a learning effect: on a no-KR retention test, the three groups did not differ.

Several researchers have reported results indicating that activity during the post-KR interval hinders learning. Of these, only those by Benedetti and McCullagh (1987) and Swinnen (1990, experiment 3) included appropriate tests for learning. In both of these experiments, the interfering activity was a cognitive activity. Participants in the experiment by Benedetti and McCullagh engaged in a mathematics problem-solving task, whereas those in the experiment by Swinnen guessed the movement-time error of a lever movement the experimenter made during the post-KR interval.

Only one experiment (Magill, 1988) has demonstrated that beneficial learning effects can result from activity in the post-KR interval. Participants learned a two-component arm movement in which each component had its own criterion movement time. During the post-KR interval, one group had to learn two additional two-component movements, one group had to learn a mirror-tracing task, and a third group did not engage in activity. Results showed that the two groups that engaged in activity during the post-KR interval performed better than the no-activity group on a transfer test in which they learned a new two-component movement.

What do these different effects of activity tell us about learning processes that occur during the post-KR interval? They support the view we discussed earlier that learners engage in important planning activities during this time period. They use this planning time to take into account the discrepancy between the task-intrinsic and the augmented feedback, to determine how to execute the next attempt at performing the skill. Much of this planning seems to require cognitive activity; we see this in the experiments showing that engaging in attention-demanding cognitive problem-solving activity during this interval hinders learning.

For many years, the view was that augmented feedback should be given during or after every practice trial (i.e., 100 percent frequency), because no learning occurred on trials without augmented feedback. However, beginning with the influential review and evaluation of the KR literature by Salmoni, Schmidt, and Walter (1984) and continuing to the present time, this traditional view has been revised as researchers have provided evidence that is contrary to predictions of that viewpoint.

The Reduced Frequency Benefit

Sufficient research evidence has now accumulated for us to say confidently that the optimal frequency for giving augmented feedback is not 100 percent. The most influential evidence to support this conclusion was an experiment by Winstein and Schmidt (1990). They had participants practice producing the complex movement pattern shown in the top panel of figure 15.10 by moving a lever on a tabletop to manipulate a cursor on a computer monitor. During the two days of practice, participants received KR after either 100 percent or 50 percent of the trials. For the 50 percent condition, the experimenters used a fading technique in which they systematically reduced the KR frequency; they provided KR after each of the first twenty-two trials of each day, then had participants perform eight trials with no KR, then systematically reduced the frequency from eight to two trials for the remaining eight-trial blocks each day. The results of this experiment are presented in the bottom panel of figure 15.10. In a no-KR retention test given one day later, the faded 50 percent frequency condition led to better retention performance than the 100 percent condition produced. In fact, people who had received KR after every practice trial showed retention test performance at a level resembling that of their first day of practice.

The Winstein and Schmidt (1990) study has generated a great deal of research since it was published. The research has focused on two predominant themes: providing additional empirical support for the reduced frequency benefit, and determining whether or not an optimal frequency exists to enhance skill learning. Two interesting conclusions have resulted from these research efforts. First, although a reduced frequency of augmented feedback can benefit motor skill learning, it may not benefit the learning of all motor skills. Second, an optimal relative frequency appears to be specific to the skill being learned and likely also to the skill level of the learner (e.g., Guadagnoli & Lee, 2004).

Theoretical Implications of the Frequency Effect

The challenge for those interested in developing motor learning theory is to establish why giving augmented feedback less than 100 percent of the time during practice is better for skill learning. One possible reason is that when people receive augmented feedback after every trial, they eventually experience an attention-capacity "overload." After several trials, the cumulative effect is that there is more information available than the person can handle.

A more likely possibility is that giving augmented feedback on every trial leads to engaging the learner in a fundamentally different type of learning processing than he or she would experience if it were not given on every trial. Schmidt and his colleagues (e.g., Salmoni, Schmidt, & Walter, 1984; Winstein & Schmidt, 1990) proposed this view, which they called the guidance hypothesis. According to this hypothesis, if the learner receives augmented feedback on every trial (i.e., at 100 percent frequency), then it will effectively "guide" the learner to perform the movement correctly. However, there is a negative aspect to this guidance process. By using augmented feedback as the guidance source, the learner develops a dependency on the availability of augmented feedback so that when he or she must perform the skill without it, performance suffers. In effect, augmented feedback becomes a crutch for the learner that is essential for performing the skill.

The hypothesis further proposes that receiving augmented feedback less frequently during practice encourages the learner to engage in more beneficial learning strategies during practice. For example, active cognitive and movement problem-solving activities increase during trials with no augmented feedback and learners are more likely to process task-intrinsic sources of feedback (Anderson, Magill, Sekiya, & Ryan, 2005). The learner does not become dependent on the availability of augmented feedback and therefore can perform the skill well, even in its absence. (For an extensive review of research investigating the guidance hypothesis and an experiment that provides support for it, see Maslovat, Brunke, Chua, & Franks, 2009.)

Now that we have established that it is generally more effective to provide augmented feedback less frequently than on every practice trial, a question that remains concerns how to implement ways that reduce frequency. The fading technique, described in the previous section, is one useful approach to reducing frequency. But several other techniques are effective as well.

Performance-Based Bandwidths

Earlier in this chapter we discussed a strategy for presenting augmented feedback that involved providing it only when a person's performance error was larger than a predetermined amount. We referred to this strategy as giving augmented feedback according to performance-based bandwidths. If we relate this bandwidth technique to the augmented feedback frequency issue, we can see how the bandwidth technique influences the frequency of presenting augmented feedback.

Lee, White, and Carnahan (1990) were the first to investigate this relationship. In their experiment they paired individual participants so that one of each pair received KR only on the trials on which the other of the pair received KR in 5 percent and 10 percent bandwidth conditions. The reason for the pairing of participants (a procedure known as "yoking") was to control for the possibility that the performance-based bandwidth benefit for learning was due to a reduced KR frequency. Thus, KR frequency was the same for each pair of participants, but the KR frequency for the participant in the bandwidth condition depended on the 5 and 10 percent criteria. Results showed that the bandwidth-based KR conditions led to better retention performance. The researchers concluded that the performance-based bandwidth technique reduces augmented feedback frequency, which is an important reason why the technique enhances learning.

Two additional experiments will serve as examples of research that provide support for the learning benefit derived from the relationship between the performance-based bandwidth technique and augmented feedback frequency. Goodwin and Meeuwsen (1995) compared 0 percent and 10 percent error bandwidth conditions with those that were systematically expanded (0-5-10-15-20 percent) and contracted (20-15-10-5-0 percent) for learning to putt a golf ball a criterion distance. The two conditions that resulted in the best retention test performance showed interesting KR frequencies. The frequencies for the 10 percent bandwidth condition reduced from 62 percent during the first twenty trials, to between 47 percent and 50 percent on the remaining trials. For the expanding bandwidth condition, KR frequencies began at 99 percent for the first twenty trials when the bandwidth was 0 percent, but then eventually reduced to 19 percent by the end of the practice trials, as the bandwidths increased in size. Lai and Shea (1999) reported similar findings for the learning of a complex spatial-temporal movement pattern. These results indicate that for the performance-based bandwidth technique it is the reduction of KR frequency during practice that is important to improved learning.

From an instructional perspective, the bandwidth technique provides a useful means of individualizing the systematic reduction of the frequency of augmented feedback in practice situations. The bandwidth technique gives the practitioner a specific guideline for when to provide augmented feedback that encourages the learner to engage in important learning strategies. And because the bandwidth is related to individual performance, the learner can engage in these strategies at his or her own rate.

Self-Selected Frequency

Another technique that bases the frequency of augmented feedback on the individual involves the learner receiving augmented feedback only when he or she asks for it. The learning benefits derived from this approach appear to result from the learner participating more actively in determining characteristics of the practice conditions by self-regulating the presentation of augmented feedback. An experiment by Janelle, Kim, and Singer (1995) provided initial evidence that this strategy can enhance the learning of motor skills. College students practiced an underhand golf ball toss to a target on the ground. The students received KP about ball force, ball loft, and arm swing during practice. Compared to groups that received KP according to experimenter-determined frequencies (all of which received it less frequently than on every trial), the participants who controlled KP frequency themselves performed more accurately on the retention test.

Janelle substantiated and extended these results in a later study in which videotape replay was a source of augmented feedback in addition to verbal KP (Janelle, Barba, Frehlich, Tennant, & Cauraugh, 1997). Participants in the self-regulated group controlled the augmented feedback schedule by requesting KP at will during 200 practice trials. An important characteristic of this experiment was that individual participants in another group were paired (i.e., yoked) to participants in the self-regulated condition to receive KP on the same trials, but without requesting it. The importance of this yoked condition was to control for the possibility that the effect of self-regulation of augmented feedback is due only to reduced frequency. Results showed that participants in the self-regulated condition learned the throwing accuracy task with more accuracy and better throwing technique than participants in the other KP and yoked conditions.

In terms of augmented feedback frequency, it is interesting to note that participants in the self-controlled condition requested KP on only 7 percent of the practice trials in the Janelle et al. (1995) experiment, and on only 11 percent in the Janelle et al. (1997) experiment. These low frequencies indicate that there is some relationship between the self-controlled procedure and the reduced relative frequency of augmented feedback. However, because people in the self-controlled conditions in both experiments performed better on retention tests than those in the frequency-yoked conditions, the benefit of the self-controlled situation is more than a simple frequency effect.

Why do beginners ask for feedback from an instructor? In one of the first experiments to address this question Chiviakowsky and Wulf (2002) asked participants in an experiment involving the learning of a sequential timing task to respond to the question "When/why did you ask for feedback?" Two-thirds of the participants answered that it was after what they considered to be a good trial (i.e., a trial during which they thought their performance was relatively successful). None indicated they asked for feedback after what they thought was a bad trial. Similar findings have been reported by these researchers in two additional experiments (Chiviakowsky & Wulf, 2005, 2007). Additionally, in a study by Laughlin et al. (2015), beginners practiced a 3-ball cascade juggling task. The participants who asked for KR, typically asked after good trials primarily to confirm their success. Those who asked for KP, asked after both good and bad trials, primarily to identify technique flaws.

Research showing that learning is enhanced when learners can select when they want augmented feedback, and the fact that this selection typically occurs after trials that are thought to be relatively successful, provides interesting insight into the role of augmented feedback in skill learning. In general these results indicate the importance of augmented feedback as a source of information to confirm a learner's subjective evaluation of his or her performance. Two points are especially relevant from these findings. First, the use of augmented feedback in this way allows beginners to engage in their own problem-solving strategies as they learn the skill. Second, these results provide excellent evidence that learners use augmented feedback as a source of motivation to continue to practice. When they perform a relatively successful trial they ask for augmented feedback to reinforce their own subjective evaluation of their performance, which encourages them to continue to practice the skill. (For a more in-depth discussion of why receiving augmented feedback after good trials benefits learning, see Carter, Smith, & Ste-Marie, 2016 and Chiviakowsky & Wulf, 2007.)

Summary and Averaged Augmented Feedback

Another way to reduce the frequency of augmented feedback presentations is to give a listing of performance-related information after a certain number of practice trials. This technique, which is known as summary augmented feedback, reduces the presentation frequency of augmented feedback while providing the same amount of information as if it were given after every trial.

The summary technique could be advantageous in several types of skill learning situations. For example, suppose that a therapy patient must do a series of ten leg extensions in relatively rapid succession. To give augmented feedback after every extension may not be possible, if time limits restrict access to performance information after each attempt. A summary of all ten attempts could help overcome this limitation. Or suppose that a person is practicing a shooting skill for which he or she cannot see the target because of the distance involved. Efficiency of practice could be increased if that person did not receive augmented feedback after each shot, but received information about each shot after every ten shots.

The first study to generate a great deal of interest in the summary technique was a laboratory-based experiment by Schmidt, Young, Swinnen, and Shapiro (1989). The task involved moving a lever along a trackway to achieve a goal movement time. During the practice trials, participants received KR after every trial or in summary form after five, ten, or fifteen trials. The results of this experiment (see figure 15.11) showed very little difference between the conditions during practice and on a retention test given 10 min after practice. But on a retention test two days later, the group that had received KR after every trial performed the worst, whereas the group that had received summary KR after every fifteen trials performed the best.

In an experiment involving the learning of target shooting with rifles, Boyce (1991) provided one group with KP after every shot, and another group with KP about every shot after each fifth shot. The results showed no difference in eventual shooting performance by these groups. Although the summary method did not yield better performance than the method giving KP after every shot, its effectiveness as an instructional technique was established, because it was just as effective for improving performance as the other method.

Numerous studies have provided evidence supporting the benefit of the summary technique for motor skill learning (e.g., Guay, Salmoni, & Lajoie, 1999; Herbert, Heiss, & Basso, 2008; Schmidt, Lange, & Young, 1990; Wright, Snowden, & Willoughby, 1990). However, a very practical question remains to be answered: What is the optimal number of performance attempts, or practice trials, to include in a summary augmented feedback statement? Although there have been several attempts to determine whether a specific number of trials is optimal, the results have been equivocal. Two answers appear reasonable at the present time on the basis of results of research that has investigated this question.

First, Sidaway, Moore, and Schoenfelder-Zohdi (1991) concluded that the positive effects of the summary technique are not due to the number of trials summarized, and stated that their results argue against the notion of an "optimal summary length." Instead, they argued that the summary effect is related either to the reduced frequency of presenting augmented feedback or to the trials delay involved in presenting augmented feedback using the summary technique (note that the summary technique has characteristics similar to the trials-delay procedure, which we discussed earlier in this chapter).

An alternative answer is one we have seen related to many of the issues discussed in this book. The "optimal" summary length may be specific to the skill being learned. Guadagnoli, Dornier, and Tandy (1996) provided evidence for this possibility by showing that longer summaries are better for the learning of simple skills, whereas shorter summaries lead to better learning of more complex skills.

One of the possible strategies people may use when presented with a listing of augmented feedback is to estimate an average for the series of trials that the summary includes. The research literature provides some evidence that this may occur. In experiments in which the learner receives the average score for all the trials in a series, the results have shown that this procedure leads to better learning than presenting augmented feedback after every trial (Young & Schmidt, 1992), and no better or worse than after every trial or after every third trial (Wulf & Schmidt, 1996). But when compared to the summary technique, no differences are found in terms of their influence on skill learning (Guay, Salmoni, & Lajoie, 1999; Weeks & Sherwood, 1994; Yao, Fischman, & Wang, 1994).

Finally, why are these two feedback presentation methods effective? Their effectiveness is undoubtedly due to the same factors that lead to the benefit of reducing augmented feedback frequency, as explained by the guidance hypothesis. During practice trials on which they receive no augmented feedback, people engage in beneficial learning activities that are not characteristic of people who receive augmented feedback after every trial.

• Augmented feedback is performance-related feedback that is provided by an external source; it adds to or enhances task-intrinsic feedback, which is performance-related feedback directly available to the sensory system during the performance of a skill.

• Two types of augmented feedback are distinguished based on the aspect of performance to which the information refers:

  • Knowledge of results (KR) refers to the outcome of an attempt to perform a skill.

  • Knowledge of performance (KP) refers to the movement-related characteristics associated with an attempt to perform a skill.

• Augmented feedback plays two roles in the skill learning process:

  • It facilitates achievement of the task goal.

  • It motivates the learner to continue to strive toward the achievement of a goal.

• The need for augmented feedback for skill learning can be described in four different ways:

  • It can be essential for skill learning.

  • It may not be essential for skill learning.

  • It can enhance skill learning beyond what is possible without it.

  • It can hinder skill learning.

• Augmented feedback content issues include the following:

  • Should the information conveyed to the learner refer to the errors made or to those aspects of the performance that were correct?

  • Should the augmented feedback be KR or KP?

  • Should the augmented feedback be quantitative or qualitative?

  • Should the augmented feedback be based on the size of the error(s) and/or number of errors?

  • What is the effect of erroneous augmented feedback on skill learning?

• KP can be presented to the learner in several different forms:

  • Verbal KP, which can provide either descriptive or prescriptive information.

  • Video replays of skill performances.

  • Movement kinematics and kinetics associated with an attempt to perform a skill.

  • Biofeedback.

• In addition to giving augmented feedback after a person has completed a trial, or after the performance of a skill (i.e., terminal augmented feedback), it can be presented during the performance (i.e., concurrent augmented feedback).

• Concurrent augmented feedback can have negative and positive effects on skill learning.

• Two intervals of time associated with terminal augmented feedback are the KR-delay interval and the post-KR interval. Both require a minimum length of time, although a maximum length has not been determined. Engaging in activity during these intervals can hinder, benefit, or have no effect on skill learning.

• Research indicates that the optimal frequency for giving augmented feedback is less than on every practice trial. The guidance hypothesis represents the most commonly held view for explaining the learning benefit of a reduced frequency.

• Several techniques will reduce augmented feedback frequency:

  • Performance-based bandwidths.

  • Performer-selected frequency.

  • Summary and averaged augmented feedback.

• Evaluate the need for KR or KP in any skill instruction situation in terms of the type of augmented feedback that would most effectively facilitate learning the skill.

• More specific or technologically sophisticated augmented feedback is not necessarily better. Beginners need feedback that will help them make a "ballpark" approximation of the movements they need to make to achieve the action goal.

• Augmented feedback that is a combination of error-correction information and information about what was done correctly can be helpful for skill acquisition and motivation to continue to try to achieve the goal of the task.

• Determine the verbal KP to give according to the most critical error made during a practice attempt. Identify this error on the basis of an analysis of the skill's component parts and a prioritized list of the importance of each part for achieving the action goal.

• Prescriptive verbal KP is better than descriptive verbal KP for beginners.

• Video replays can be effective as augmented feedback for beginners when you point out errors and provide information about how to correct them. The decision to provide this type of information for more skilled individuals can be based on the individual's choice.

• Computer-generated displays of the kinematics of a skill performance will be more effective for learners who are at a more advanced stage of learning than a beginner.

• Biofeedback can facilitate skill learning when it provides information people can use to alter movements and when they do not become dependent on its availability.

• Do not feel compelled to give augmented feedback after every practice attempt. When you do not give augmented feedback, you provide opportunities for people to determine what their own sensory feedback tells them about performing the skill they are learning.

• The performance-bandwidth strategy of providing augmented feedback can be especially useful when instructing groups of individuals where it is difficult to interact with each person individually on every performance attempt.

• Allow people you are working with to determine when they would like to receive KR or KP.

• On occasion, ask the people you are working with to tell you what movement errors they made and how they should correct them before you give them this information.

1. (a) Describe the two general types of performance-related feedback a person can receive during or after performing a motor skill. In your description, indicate the characteristic that differentiates the two. (b) Discuss why the distinction between these two types of feedback is important.

2. What are the two types of information referred to by the terms KR and KP? Give two examples of each.

3. Describe skill learning conditions where augmented feedback would (a) be necessary for learning, (b) not be necessary for learning, and (c) not be necessary for learning but would enhance learning beyond what would occur without it.

4. (a) How do quantitative and qualitative augmented feedback differ, and how do they influence the learning of motor skills? (b) Describe how you would use these two forms of augmented feedback in a motor skill learning situation.

5. Describe a situation in which you would use video replay as a form of augmented feedback to (a) help a beginner learn a new skill, (b) help a skilled person correct a performance problem. Indicate why the video replay would facilitate learning for each situation.

6. Describe a situation in which you would use kinematic information as augmented feedback to help someone learn a motor skill and explain why you would use it.

7. Describe a skill learning situation in which you would use some form of biofeedback. Indicate how you would use it, and why you would expect it to facilitate the learning of the skill.

8. What is the difference between concurrent and terminal augmented feedback? Give two examples of each.

9. (a) What are two types of activity during the KR-delay interval that have been shown to benefit skill learning? (b) Why does this benefit occur?

10. (a) What seems to be the most appropriate conclusion to draw regarding the frequency with which an instructor should give augmented feedback during learning? (b) How does the guidance hypothesis relate to the issue of augmented feedback frequency?

11. Describe a skill learning situation in which (a) giving summary augmented feedback would be a beneficial technique and (b) using the self-selected frequency strategy would be beneficial.

• To read about a popular video-based performance analysis program that can be used to provide augmented feedback in sports, physical education, and physical rehabilitation settings, go to http://www.dartfish.com.

• To read a summary of a study, funded by the National Athletic Trainers Association (NATA) Foundation, which investigated the use of augmented feedback to help people decrease impact forces of a jump landing, go to http://natafoundation.org/wp-content/uploads/Onate.pdf. This report was published in the Supplement to the Journal of Athletic Training, April–June, 2000, Vol. 38, No. 2, pp. 19–20, and was included in a presentation at the 2003 NATA Annual Meeting in St. Louis, Missouri.

• To read an article for rowing coaches about the use of augmented feedback to change a rower's technique, go to http://www.brianmac.demon.co.uk/articles/scni19a4.htm.

• To see a video on different types of augmented feedback that might be provided to rowers go to http://www.youtube.com/watch?v=7H_3c2kRl2c

• To read a summary of the development and use of a training system, which involves a virtual environment and augmented feedback, designed to assist in the physical rehabilitation of people with neurological impairments, go to http://web.mit. edu/bcs/bizzilab/members/holden/. Included on this page is a list of articles the researcher has published concerning the training system.

• To read about other projects in this laboratory at M.I.T., go to http://web.mit.edu/bcs/bizzilab/.