Introduction

A golf swing is a highly complex motor skill that requires coordination of multiple segments of the body (Knight, 2004). To maximize club head speed, golfers must make interactive movements such as proper kinematic sequence (Crews & Lutz, 2007; Neal et al., 2008; Vena et al., 2011). There have been several previous studies employing various biomechanical analyses of swings in order to improve golf performance (An et al., 2013; Hume et al., 2005; Okuda et al., 2010; Tinmark et al., 2010). However, considering the comparatively limited attentional capacity demands of the human brain, verbally demonstrating the complex information gleaned from those analyses in golf swing lessons is exceedingly difficult. In fact, application of complex and detailed information can decrease motor performance and learning effects because it can overload cognitive processing (An et al., 2013; Liao & Master, 2001; Masters et al., 2002). By contrast, application of more efficient information induces automated motion and facilitates the motor learning process (Schmidt & Lee, 2013).

Nevertheless, much of teaching in sports education continues to rely on the “cramming method”, wherein teachers try to explain complex and detailed movement information in learning (Janelle et al., 1997; Hebron, 2017). However, such methods can not only overload cognitive processing but also encourage excessive dependency on the instructors, degrading learners’ ability to memorize information (e.g. encoding, rehearsal, and recognition; Ross et al., 2008). Therefore, many motor learning specialists have emphasized the importance of a self-learning environment involving the use of different feedback methods (i.e. faded [progressively reducing feedback], self-controlled [giving feedback only when asked], and bandwidth [giving feedback only outside of set error tolerance limits]). Patterson & Lee (2008) emphasized the importance of self-error detection and correction in the motor learning process: “It isn’t about how much you practice. It is about how you practice.”

Accordingly, we suggest that reactive neuromuscular training (RNT) could be effectively applied in the self-learning environment. RNT is a neurological learning technique centred on “feeding the mistake”. It exaggerates performers’ mistakes in order to create an unstable environment using training aids such as elastic tubing bands (Voight & Cook, 1996). When incorrect movements are exaggerated via external perturbation, performers can better perceive proprioceptor feedback of the incorrect posture. This improved awareness of mistakes regarding movement perturbation in turn, which guides the performers towards a self-learning environment wherein they can stabilize their movements and align their joints themselves (Pittman, 2013). RNT requires minimal linguistic advice for movement correction and enhances the performer’s self-error detection and correction ability in motor learning. For example, when attempting to correct knee valgus, RNT would involve exaggerating the knee valgus by using an elastic band rather than verbal or visual feedback (Cook et al., 1999). RNT is a learning method that promotes proprioception and helps learners embody the movements required to acquire correct coordination structures. According to Fitts and Posner’s cognitive learning stages (Cognitive – Associative – Autonomous), the cognitive overload and limited problem-solving ability seen in the early stages of learning are well explained (Fitts & Posner, 1967). In this context, we can expect RNT to be effective for motor skill acquisition.

In previous studies, RNT was applied to the rehabilitation of patients with orthopaedic or neurological lesions. Female college basketball players who had anterior cruciate ligament surgery were found to have recovered knee joint stability and improved asymmetric leg strength via RNT (Cook et al., 1999). Pittman (2013) also verified the benefits of RNT through three-dimensional (3D) motion analysis. He found that knee valgus, which causes problems with the anterior cruciate ligament, showed greater improvement after RNT than after traditional resistance training. Moreover, Seada et al. (2013) investigated the benefits of RNT for walking performance among patients with Parkinson’s disease using 3D motion analysis and electromyography. They found that the RNT group – who received training in their balance with unexpected perturbation – showed dramatically improved walking performance, including walking stride, speed, fall rate, and muscle activity. Other RNT studies that have been conducted on patients with orthopaedic or neurological lesions revealed similar results. However, there have been very few studies investigating the effect of RNT on motor skill learning, such as hand grenade throw (An et al., 2016), golf skills (An & Lee, 2018; An, 2024). One of the few existing studies on simple motor skills is related to the “overarm throw” (practice hand grenade). In that study, both performance and qualitative movement analysis were conducted, and significant effects of RNT group was observed in the accuracy of performance as well as in the posture of the overarm throw (arm, torso, and leg movements).

When applying RNT to motor skill learning, it is necessary to verify a performer’s error tendencies. One of the most common swing faults among novice golfers relates to the “head or body sway” movement. More specifically, novice golfers tend to use allocentric coordination patterns, moving their heads and clubheads together in the same direction (Lee et al., 2008); therefore, kinematic sequence of deceleration timing across the three segments (pelvis, thorax, wrist) tend to appear towards the end of the downswing, rather than the mid- downswing (Putnam, 1993; Crews & Lutz, 2007). Thus, in this study, we tried to apply RNT to novice golfers who tend to commit this fault using the inertial overloading technique. Inertial overloading is based on Newton’s second law (F = M × A), that is, the vector sum of the force (F) is the product of the change in mass (M) and velocity (V). We assumed that the inertia overloading technique would exaggerate the performer’s head or body sway movement, thus creating more body perturbation and instability. The purpose of this study was to investigate the learning effect of RNT (i.e. the inertia overloading task) on golf performance and kinematic sequence for a full swing of a 7-iron golf club.

Methods

Procedure

The experiment proceeded in three stages (pre-test, practice phase, post-test). Prior to the pre-test, the basic golf swing instruction and mechanism were demonstrated by a professional golfer to each participant (e.g., standard grip, stance, ball position, body alignment, weight transfer, swing plane and swing arc). The experiment task was performing a 7-iron golf full swing to a 5m of golf target net (indoor golf swing room). To determine the homogeneity of golf skills among participants, 10 trials of 7-iron golf full swings were measured in the pre-test. The practice phase proceeded over the 4-week period (60 practice trials twice per week). During the practice phase, the RNT group had practiced a 7-iron golf swing with RNT technique (inertia overloading task). Control group had only practiced a 7-iron golf swing without RNT technique. The RNT groups performed the inertia loading task with a 5kg of clubbell. The inertial loading tasks were organized according to the CST clubbell swing manual (i.e. front swing, side swing, hammer swing; 3 sets of 10 repetitions). The front swing action occurred within the sagittal plane of the body, while the side and hammer swings occurred within the frontal plane. These three basic swing actions were selected in order to cause perturbations through exaggeration of the body sway in 3D space. To avoid fatigue, participants were given a three-to-five minute break between training tasks. The experiment procedures are shown in Table 1.

View original

Table 1

Experimental procedure

Weeks	RNT group	Control group
0	Pre-test (10 trials of 7 iron golf full swing)
1	7-Iron golf full swing [3sets of 20 trials] + RNT task [10 front swings + 10 side swings + 10 hammer swings) × 3 sets]	7-Iron golf full swing [3sets of 20 trials]
2
3
4
5	Post-Test (10 trials of 7 iron golf full swing)

To examine the learning effect of RNT, we assessed a post-test after the four-weeks of practice phase. To examine the golf performance, we had participants hit 10 golf balls using a 7 iron as far as possible, while aiming at the target net. All participants wore the K-vest during the test. Velcro was used to minimize the movement of the sensors during the full swing. Before we began data collection, we had participants perform several practice swings to make them comfortable with the K-vest. After calibrating the inertia sensors (by having subjects stand upright with a static posture and hold this posture for two to three seconds), each participant performed 10 full swings with a 7 iron. The experimental setup is shown Figure 1.

View original

Figure 1

Experimental setup.

Results

Carry Distance

The results for carry distance are shown in Figure 2. As is evident, the RNT group outperformed the control group after the 4-week of practice phase. Furthermore, the main effect of the group was significant, F (1, 22) = 7.782, p < .05. The main effect of the test period was also significant, F (1, 22) = 45.384, p < .001. Additionally, the interaction of the group and test period was significant, F (1, 22) = 13.837, p < .01. Accordingly, one-way ANOVA was conducted to examine the simple main effect at each level. In the pre-test, there was no significant difference between groups, F (1, 22) =.558, p > .05. However, there was a significant difference in the post-test, F (1, 22) = 15.779, p < .001.

View original

Figure 2

Carry Distance of 7-iron golf full swing.

Maximum Angular Velocity

The maximum angular velocity of the pelvis tended to increase in RNT and Control groups, and the main effect of the test period was significant, F (1, 22) = 12.575, p < .01. The interaction of group and test period was also significant, F (1, 22) = 8.966, p < .01. However, the main effect of the group was not significant, F (1, 22) = 3.257, p > .05. One-way ANOVA was conducted to examine the simple main effect at each level. In the pre-test, there was no significant difference between groups, F (1, 22) =.136, p > .05. However, there was significant difference in the post-test, F (1, 22) = 7.740, p < .05 (Figure 3a).

View original

Figure 3

Maximum angular velocity (a: pelvis, b: trunk, c: wrist).

The maximum angular velocity of the thorax is shown in Figure 3b. The interaction of the group and test period, F (1, 22) = 6.037, p <. 05, and the main effect of test period, F (1, 22) = 40.681, p < .001 were significant. However, the group main effect was not significant, F (1, 22) = .570, p > .05. One-way ANOVA was conducted to examine the simple main effect at each level. In the pre-test, there was no significant difference between groups, F (1, 22) =.117, p > .05. Moreover, there was no significant difference in the post-test, F (1, 22) = 2.379, p > .05.

Finally, the maximum angular velocity of the wrist tended to increase in both groups, but the RNT group showed a greater increase (Figure 3c). The interaction of the group and test period, F (1, 22) = 19.751, p <. 01, and the main effect of test period, F (1, 22) = 50.830, p < .001 were significant. However, the main effect of the group was not significant, F (1, 22) = 1.814, p > .05. One-way ANOVA was conducted to examine the simple main effect at each level. In the pre-test, there was no significant difference between groups, F (1, 22) =.309, p > .05. However, there was significant difference in the post-test, F (1, 22) = 8.324, p < .01.

Deceleration Timing of Maximum Angular Velocity

The deceleration timing of the pelvis showed a dramatic forward shift in the RNT group compared with the Control group (Figure 4a). The main effect of the group was significant, F (1, 22) = 7.387, p < .05, and the main effect of test period was significant, F (1, 22) = 33.778, p < .001. However, the interaction of the group and test period was only borderline significant, F (1, 22) = 4.165, p = .053. One-way ANOVA was conducted to examine the simple main effect at each level. In the pre-test, there was no significant difference between groups, F (1, 22) =.005, p > .05. However, there was a significant difference in the post-test, F (1, 22) = 8.399, p < .01.

View original

Figure 4

Deceleration timing of rotation (a: pelvis, b: trunk, c: wrist).

The deceleration timing of the thorax is shown in Figure 4b. The interaction of the group and test period, F (1, 22) = 4.658, p < .05, and the main effect of the test period were significant, F (1, 22) = 5.752, p < .05. However, the main effect of group was not significant, F (1, 22) = 3.203, p > .05. One-way ANOVA was conducted to examine the simple main effect at each level. In the pre-test, there was no significant difference between groups, F (1, 22) =.095, p > .05. However, there was a significant difference in the post-test, F (1, 22) = 7.471, p < .05.

Figure 4c shows the wrist deceleration timing. Neither the main effect of the group, F (1, 22) = .303, p > .05, nor the interaction of group and test period, F (1, 22) = .370, p > .05 was not significant. Moreover, the main effect of test period, F (1, 22) = .311, p > .05, was not significant.

Discussion

The results of the study indicated that the RNT group showed significant enhancements in the deceleration timing of the proximal segments and the kinematic sequence. As a result, the RNT group outperformed the Control group in carry distance as well as the maximum angular velocity of the pelvis and wrist. In this study, RNT involved overloading inertia to exaggerate novice golfers’ head or body sway during the swing.

To understand why novice golfers move their heads with their golf swings, we must focus on the coordination strategy of different expertise. The first concept of such a coordination strategy is the “proximal-to-distal sequence” (PDS) pattern (Putnam, 1993; Hume et al., 2005; Crews & Lutz, 2007; Neal et al., 2008; Tinmark et al., 2010). Putnam (1993) emphasized that the PDS is a key element of maximizing inertia force by generating multi-joint interaction torque. The PDS pattern can be analysed in terms of the timing of maximum angular velocity. Specifically, for an expert’s golf swing, the maximum angular velocity of the three segments are in a PDS pattern (pelvis-thorax-wrist), meaning that the proximal segment begins to decelerate when the distal segments accelerate. Moreover, the timing of the proximal segments’ maximum angular velocity occurs in the middle of the downswing (at around 60% of the downswing phase). These results indicate that experts show a release of degrees of freedom as well as greater multi-jointed interaction torque (Dunn & Putnam, 1988). However, novice golfers’ kinematic sequence is more varied. Owing to the freezing degree of freedom and reverse sequence pattern, the timing of the maximum angular velocity tends to be delayed towards the end of the downswing in a “distal-to-proximal” sequence pattern (e.g. thorax-pelvis-wrist, wrist-thoraxpelvis). Therefore, the upper body (e.g. head) sways more in the golf swing.

The most meaningful finding of this study is the deceleration sequence timing of the pelvis and trunk. Generally, wrist deceleration timing occurs near the point of impact (between 95–100%), which may explain why no significant differences were found between groups in this area. In the pre-test of this study, the proximal segments’ maximum angular velocity (pelvis & trunk) appeared mostly towards the end of the downswing, at around 95–100%. However, over the 4-week practice period, the RNT group showed a dramatic timing shift forward in the downswing (pelvis: 80~90%, trunk: 85~95%). This can be interpreted as evidence of learning a more effective coordination pattern (PDS; Proximal to Distal Sequence), as required in rotational sports skills such as the golf swing. On the other hand, the Control group continued to show a similar deceleration timing sequences as the pre-test, such as simultaneous rotation timing. As a result, the RNT group outperformed the Control group in carry distance. Moreover, the maximum angular velocity of the pelvis and wrist showed significant improvements in the RNT group. These results imply that the RNT group learned more effective coordination in order to maximize club head speed by maintaining a centre arc in their golf swing. In previous studies, the temporal relations of the proximal-to-distal sequence were also emphasized as a critical factor related to accuracy performance (Furuya & Kinoshita, 2007; Hirashima et al., 2007). This is because early deceleration of the proximal segments would generate better multi-joint interaction (Tinmark et al., 2010).

The second concept of the coordination strategy is an allocentric and egocentric coordination pattern. Lee et al. (2008) examined the coordination of the head and putter movement in expert and novice golfers and demonstrated that the motionless head movement observed in expert golfers was because they followed an egocentric coordination pattern: that is, their head and putter movements were tightly coupled but in opposite directions. Conversely, novice golfers used an allocentric coordination pattern: that is, they moved their head and putter together in the same direction. Later, Lee et al. (2008) demonstrated that experts’ egocentric coordination pattern was the optimal strategy for stabilizing body movement with enhanced power during a golf swing. These results explain why the deceleration timing of proximal segments appears towards the middle of the downswing in expert golfers and towards the end of the downswing in less skilled golfers with simultaneous rotation strategy.

Based on this evidence, it is possible that novice golfers can learn an egocentric coordination pattern (i.e. a counter-balancing strategy or an early deceleration of the proximal segments) after 4 weeks of RNT. Newell (1991) demonstrated the “search strategy” concept in motor skill acquisition. Specifically, the search strategy refers to an individual’s process of learning optimal motor solutions for their perceptualmotor workspace (Newell et al., 1989) – that is, learning how to coordinate one’s perceptual environment with the action environment in a way that is consistent with the task constraints (Newell, 1991). RNT provides a self-learning environment, and thus forces performers to search for optimized coordination for their unstable environment through proprioceptive feedback. Shadmehr and Mussa-Ivaldi’s (1994) robot arm stretch experiment indirectly explains how optimal coordination can be acquired in a self-learning environment via a search strategy. In their task, which involved reaching a targeted point with a hand, a robotic device was used to disturb the reaching action. In the early stage of training, performance errors increased notably, but the stability of hand motion gradually improved with repetition. This result implies that the subjects learned a compensation strategy and optimal movement in an unstable environment through a “search strategy”. Moreover, Kelso (1984) demonstrated a phase transition into more stable and functional attractors by increasing the speed of a bi-manual finger coordination task. The RNT group might have experienced a phase transition to obtain a more stable attractor.

Indoctrination teaching methods, such as repetitive verbal feedback or informing learners of detailed movements for each performance, might delay learning by overloading learners’ information processing and cultivate dependence on instructors. To speed-up learning, performers should detect and solve problems that they encounter with their performance by themselves. Creative instructors should be able to design effective RNT solutions that induce a self-learning environment. As RNT remains a new approach in motor skill learning studies, we expect future research to conceive applications for RNT in various motor learning situations. In addition to the inertial overload approach using weight, RNT can also be applied through the principle of elastic resistance using tools such as TheraBands (An, 2024).