Stabilizing Joint Angle Velocity Contributes to Motor Learning in Percussion

Abstract

This study quantified movement changes in the upper-limb joints during motor learning in percussion performance. Eleven percussionists practiced a novel excerpt on 3 consecutive training days, followed 1 week later by a retention assessment. Motion capture technology quantified the orientation and angular velocity of the shoulders, elbows, and wrists. To manipulate task difficulty, and thus the behavioral demands on upper-limb movements, participants learned the test excerpt in fast and slow tempo conditions. Given previous studies examining motor performance in percussion, it was hypothesized that variability in elbow and wrist angular velocity would be altered by performance tempo but would stabilize when comparing training vs. retention performances. The results showed greater abduction of the left shoulder in the slow vs. fast training condition. Variability in the velocity of right elbow and left wrist movements was lower at retention vs. training. Judges’ ratings of the performances revealed improved quality and rhythmic accuracy at retention vs. training. The findings may suggest that angular velocity mechanistically reflects motor learning in percussion performance. Such findings may be applied towards enhancing percussion training.

Playing a musical instrument requires complex coordination of the upper-limb joints to produce sound. Much like sports contexts, music performance involves adduction/abduction and flexion/extension of the upper limbs. Unlike sports contexts, however, studies of music performance have yet to elucidate kinematic mechanisms underlying the learning of musical motor skills. Further advancing the research area of applied motor learning in music may result in enhanced approaches to music education that could benefit individual learners, and the broader music community. To this end, the present study assessed how individual joint motions of the upper limbs were altered as motor learning progressed in a sample of trained percussionists.

Assessing motor learning in musical contexts has typically involved pianists as the population under study. In piano performance, the metacarpophalangeal joint is responsible for increasing and decreasing preparatory finger height at slower and fast tempi, respectively (Dalla Bella & Palmer, 2011). The engagement of joints proximal and distal to the fingers also differs between expert and novice pianists, with experts utilizing proximal joints such as the wrists and hands to reduce biomechanical efforts relative to distal joints such as the elbows and shoulders (Furuya et al., 2011). Additional work focusing on gross movement of the upper limbs (e.g., elbows, shoulders, etc.) identified the role of angular velocity on sound production. Interestingly, performing at fast compared to slow tempi elicits decreased angular velocity of the wrist, while shoulder and elbow angular velocity remained unchanged across tempi (Furuya & Kinoshita, 2008). However, understanding how deliberate training refines coordination patterns within musical contexts that involve both fine (i.e., individual finger movements) and gross (i.e., shoulder movements) motor control remains under investigated. A pitched percussion instrument such as the marimba requires gross motor control in the shoulders and elbows, while also involving fine movements of the fingers to shift the interval between mallets. Thus the marimba may offer an ideal perspective to further empirical understanding of how training enhances fine and gross motor control.

Marimba performance also requires rapid alterations in limb position. Marimba is a pitched percussion instrument wherein performers use mallets to strike tone bars and produce sound. As a result, playing the high and low notes of any excerpt involves significant rotation and movement of the upper limbs and torso, specifically along the mediolateral (i.e., moving from left to right) and anteroposterior axes (i.e., moving forwards and backwards). In addition, movements along the vertical z-axis appear critical for ensuring rhythmic accuracy within both snare drum and marimba contexts (e.g., Dahl, 2000, 2004, 2011; Loria et al., 2022a). The complexity of marimba performance is further compounded by the fact that musicians must exhibit high levels of fine motor control to manipulate two mallets in each hand independently to achieve desired aural outcomes (e.g., expressive gestures; see Schutz & Lipscomb, 2007). Therefore, marimba performance requires a high degree of motor control across distal and proximal joints to control mallet movements and produce sounds (Moersch, 2016). Much like the piano literature cited above, previous work has highlighted the importance of elbow and wrist control during performances.

Elbow and wrist movements mechanistically impact sound production in marimba. During early phases of motor learning, percussionists must first learn how to coordinate individual joint motions efficiently when playing, thus addressing the degrees of freedom (see Bernstein, 1967). The degrees of freedom problem may be partially solved by clustering movements of the elbow and wrist joints along the mediolateral and anteroposterior axes during a performance (i.e., Loria et al., 2021). Importantly, the difficulty of a specific performance can vary in relation to the tempo of the musical piece. Previous work has also shown greater variability in elbow and wrist velocity associated with fast (i.e., difficult) vs. slow (i.e., easy) tempi (i.e., Loria et al., 2022b). Regarding motor learning, the elbow joint may play a key role, given previous reports in marimba performance showing reduced variability in elbow range of motion as time spent training increased (i.e., Loria et al., 2022a). However, there is little empirical data regarding how individual joint motions and angular velocity control are altered when learning music at different tempi (i.e., difficulty) to enhance sound production. Because professional musicians must learn and perform repertoire across various tempi, understanding how joint-specific movements adapt during practice can provide deeper insight into how motor control strategies may transfer between different performance conditions (e.g., tempo). Developing new knowledge in this area can inform evidence-based pedagogy for percussion training, ultimately helping musicians optimize performance and meet their performance learning goals.

One goal of music education is to equip students with the proper skills to perform sound-producing movements. As such, alterations in individual joint motion occurring during the skill acquisition process could reflect attempts by a performer to improve sound production. In a recent performance study, Loria et al. (2024) randomly divided percussionists into one of two practice conditions wherein one group was trained by an instructor in person and the other group with an instructor online. The two groups performed an excerpt scored for multiple drums while attempting to increase the performance technique legato from pre- to post-training timepoints. The authors obtained recordings of the performances which were subsequently rated by expert percussionists on the overall performance quality, and legato expressivity. Irrespective of group, the kinematic results revealed greater average mallet height and elbow and wrist velocity in post- vs. pre-training performances. Interestingly, the post-training performances were rated by the judges as having better performance quality and legato expressivity relative to pre-training. This pattern of effects may highlight that changes within individual joint motions may yield improved legato production. Although we cannot directly relate improved legato production to the present study, at the very least, the findings of Loria et al. (2024) suggest that acute training can alter the movements underlying sound production. Whether movement changes would be stable and persistent over a longer period (e.g., days, one week, etc.) has yet to be determined. Developing such knowledge will likely benefit the training of musicians.

The present study assessed motor learning in marimba via individual joint motions. Participants learned a novel excerpt in two difficulty conditions (i.e., 60 bpm and 70 bpm; see Loria et al., 2022a) on three consecutive days in the laboratory (i.e., training). One week following session 3, participants returned for a retention assessment, having not practiced the excerpt in between. Auditory performance recordings were obtained for the training and retention timepoints. Data analysis focused on shoulder, elbow, and wrist angles, as well as angular velocity and variability to determine how joint movements were altered as a function of time and performance difficulty. Based on the piano and percussion literature, it was hypothesized that elbow angular velocity would initially be greater in the fast (70 bpm) vs. the slow (60 bpm) condition (see Loria et al., 2022b). However, given previous piano performance studies (e.g., Furuya & Kinoshita, 2008), wrist angular velocity would be greater in the slow (60 bpm) vs. the fast (70 bpm) condition. As training progressed, joint velocity variability was expected to stabilize, particularly at the retention assessment (see Loria et al., 2022a). Rating scores for the performances were expected to increase from training to retention, which may highlight how specific kinematic changes yield improved sound production in marimba.

Methods

Participants

Eleven percussionists from the Faculty of Music at the University of Toronto (# of females = 1) with an average of 13.6 years of percussion experience (SD = 6.2) completed the study. All participants self-reported as right-hand dominant. Participants were upper-year undergraduate and graduate students. The study was approved by the University of Toronto Research Ethics Board (Protocol #39537) and participants provided informed written consent prior to participating. The data reported below does not overlap with previously published work.

Apparatus

Participants played a Musser Deluxe Studio Grand Rosewood M245 marimba (Ludwig Musser, Elkhart, IN, United States) during the experiment. Movements of the participant's upper limbs and torso were measured using eight Vicon Vero (Vicon Motion Capture, Oxford, United Kingdom) motion capture cameras. The system's standard resolution is 2.2 megapixels (i.e., 2048 * 1088) with a camera latency of 3.6 milliseconds and spatial resolution of 0.1 mm. The system was calibrated to the lowest natural note of the marimba (A2), which served as the origin position (i.e., coordinate 0, 0, 0 across the x, y, and z movement axes).

Twenty-seven markers were affixed to the limbs of the performers in line with the upper-limb model written in Vicon BodyLanguage (e.g., Murray, 1999). For brevity, only the markers used for data analysis will be described. Markers were positioned on the left and right limbs, including on the acromion-clavicular joints (used to measure shoulder movements), the lateral epicondyle at the elbow joints (used to measure elbow movements), the midpoint of the forearms, the thumb side of the radial styloid, the little finger side of the ulnar styloid (used to measure wrist movements), and just below the third metacarpus on both hands (see Cutti et al., 2005; Murray, 1999). Considering the range of upper-limb motion during the performance, and spatial layout of the motion capture cameras, the markers were sampled at 100 Hz to ensure they could be continuously recorded throughout the performances.

Procedure

Participants practiced an excerpt from the work “Un chien dehors” by Jean-Pierre Drouet, which requires the combination of visual cues from a double staff. The excerpt (see Figure 1) required rapid alterations in mallet positions via movements of the elbows and wrists along the anteroposterior, mediolateral, and vertical movement axes. To properly execute the excerpt, participants had to find the correct mallet position for each hand and shift the rhythm pattern as the music progressed, while processing information from both staves. The rhythm shift required precise gross motor control of mallet speed at the same time as fine motor control for hand adjustments to increase or decrease the distance between mallets.

Figure 1.

The excerpt learned in the present study and a depiction of the motion capture markers used for analysis. For the markers, red = shoulders, yellow = elbows, green = wrists, and blue = hands.

Participants practiced their assigned excerpt over three consecutive days in the laboratory. Each training session was a maximum of 40 min long and participants were free to practice however they pleased, except for during specific performance assessments occurring throughout the session (see Figure 2). During their training time, which was not recorded by the experimenter, participants typically played the excerpt at a chosen tempo, made notes on the excerpt regarding sticking patterns, read over the excerpt, or focused on repeating smaller phrases within the excerpt. Participants were instructed that during the 40-min training session, the experimenter would record one performance of the excerpt using both a microphone and the motion capture system every 10 min (i.e., the performers played the entire excerpt during these recordings). The procedure for the trials where kinematic and auditory data were collected varied slightly from session to session.

Figure 2.

Organization of the training and retention sessions.

The procedure used to collect kinematic data is illustrated in Figure 2. The main focus of the experiment was to mimic traditional approaches to training that are used outside of the laboratory, which included learning the piece at a slower tempo prior to increasing tempi (i.e., difficulty; see Hartenberger, 2016; Stevens, 2000) Each session had up to three trials of the participant performing the entire excerpt. In session 1, participants were instructed that every 10 min they would be required to play the entire excerpt but could do so at any tempo for their first three trials (i.e., the first 30 min of training). To acclimate participants to the performance tempi used on the following day (i.e., session 2) they were told that regardless of the tempo used to perform the first three trials, the fourth trial had to be played at 60 bpm (i.e., 4 notes per second), at the end of session 1 (i.e., a metronome provided this tempo to participants). In session 2, participants were again free to practice at any tempo in between trials but were instructed that the first three trials would be performed at 60 bpm and the final trial was to be performed at 70 bpm (i.e., approximately 4.6 notes per second). This was done to mirror typical practice schedules wherein tempo (i.e., difficulty) is increased throughout training as learning progresses (see Hartenberger, 2016).

In session 3, participants were free to practice at any tempo in between trials but were instructed that three trials would be obtained at 70 bpm. Following session 3, participants returned to the laboratory one week later for the retention assessment. Because participants did not practice their excerpt in between session 3 and retention, 10 min of practice was provided (i.e., 5 min per tempo condition) at the start of the retention assessment. Three trials were collected for each tempo to ensure the testing session was no longer than 30 min. Due to our focus on external validity of the practice session, the order that participants performed the tempi was not counterbalanced.

Data Analysis

Data Processing

Movements of the shoulders, elbows, wrists, and hands were recorded along three movement axes. The x-axis ran from medial to lateral (i.e., left to right) along the marimba. The y-axis ran anterior to posterior (i.e., front to back) along the marimba. The z-axis ran vertically (i.e., up and down). The capture volume was calibrated to the lowest natural note of the marimba (A2). As such, the bars of the marimba were the origin position along the vertical z-axis, and the x- and y-axes originated from the midpoint of the bars. Doing so meant that movements of the upper limbs behind the marimba were negative, and movements made above/in front of the marimba were positive. Prior to data analyses, all data were signal processed using a low-pass fourth-order Butterworth Filter with a cut-off frequency of 6 Hz (i.e., Butterworth, 1930).

Data Reduction and Statistical Design

Joint angles were computed using the proprietary software provided by Vicon. Joint angles and angular velocity were calculated for the shoulders, elbows, wrists, and hands along all three movement axes (i.e., x, y, z) using the BodyLanguage Model written in Vicon Nexus (see Cutti et al., 2005; Murray, 1999). Joint angles and angular velocity were obtained for each of the three trials in the 60- and 70-bpm conditions across the training and retention timepoints. Joint angles and angular velocity were subsequently averaged across each of the three trials in the 60-training, 70-training, 60-retention, and 70-retention performances (see Figure 2). Peak angular velocity was also computed for the same timepoints and training conditions and was defined as the maximum angular velocity achieved during a trial. Peak velocity metrics are often utilized to measure feedback-driven control of goal-directed actions, with increases in peak velocity generally considered as improved motor control in a variety of movement contexts (e.g., Loria et al., 2022b; Loria et al., 2024; Muhammed et al., 2020; Schmidt et al., 1979) Joint angles, angular velocity, and peak angular velocity data were submitted to separate 2 timepoints (training vs. retention) × 2 difficulty (60-condition vs. 70-condition) repeated-measures ANOVAs for each movement axis. Of relevance were interactions between timepoints and difficulty. Tukey's HSD Test was used as post-hoc procedures when appropriate. Only significant effects were reported below.

Ratings of Performances

The auditory recordings of all the training and retention trials were judged by three experienced percussionists. They had a mean age of 30.3 years (SD = 2.2) with 16.3 years (SD = 6.6) of formal music education and 14.7 years (SD = 5.8) of percussion training. Audio recordings were first captured via a Sennheiser e845 microphone (Sennheiser, Wedemark, Germany). Recordings were subsequently compressed to reduce background noise and standardize the output. Frequencies below 50 Hz and above 8.0 kHz were filtered to reduce background noise, and compression was applied at a ratio of 2:1 at a threshold of −21.0 dB to regulate volume across recordings. After pre-processing had been completed, ID numbers were assigned to the video recordings to ensure judges were blinded to which timepoint the performance occurred. Judges rated the overall performance quality using a 6-point Likert scale of 1 (poor performance), through 3 (mediocre performance) and 6 (outstanding performance). Judges also rated rhythmic accuracy (i.e., performance timings relative to the audible metronome) of the performances using a 6-point Likert scale of 1 (very inaccurate performance), through 3 (mediocre accuracy), and 6 (outstanding accuracy). Detailed prompts given to the judges are provided in Table 1. Ratings for each participant's performance were averaged across the three judges and compared using separate 2 time (training vs. retention) × 2 difficulty (60-condition vs. 70-condition) repeated-measures ANOVAs.

Table 1.

Prompts given to the judges for rating the performances.

Prompt 1	You will be rating each recording according to its rhythmic accuracy and performance quality on a scale from 1 to 6.
Prompt 2	To evaluate performance quality you can use dynamics as indicative of musicality, but we are also interested in how participants’ melodic ability may indicate good overall performance quality. Performances with few mistakes may indicate good overall performance quality. Performances with numerous mistakes (i.e., playing the wrong note) may indicate poor performance quality.
Prompt 3	You will evaluate performance quality using a 6-point scale wherein 1 = poor performance, 3 = mediocre performance, 6 = outstanding performance.
Prompt 4	To evaluate rhythmic accuracy, have the score open while listening to compare the performance to the metronome playing in the background. Trials with uneven rhythm (i.e., either the pulse or subdivision is uneven) may indicate lower rhythmic accuracy.
Prompt 5	You will evaluate rhythmic accuracy using a 6-point scale wherein 1 = very inaccurate, 3 = mediocre accuracy, 6 = very accurate performance.
Prompt 6	Please note that the recordings you will hear do not follow the tempo indicated on the score. Participants had to perform at either 60 or 70 bpm.

Results

Performance Kinematics

Raw data illustrating the main findings are presented in Figure 3. The ANOVA for left shoulder angles along the mediolateral x-axis revealed a main effect of time, F(1, 10) = 6.9, p = .03, ${\eta }_{\mathrm{p}}^2$  = .4, and a time × difficulty interaction, F(1, 10) = 8.1, p = .02, ${\eta }_{\mathrm{p}}^2$  = .4. Post-hoc test for the interaction (HSD = 4.4°) showed greater abduction (i.e., angling the shoulder toward the body's midline) in the 60- compared to the 70-bpm conditions (60-bpm M = -10.2°, SD = 3.4, 70-bpm M = 4.7°, SD = 4.8) during training. Contrasting between training and retention revealed that shoulder abduction was reduced in retention (M = -3.1°, SD = 4.1) compared to training for the 60-bpm condition only. The ANOVA for left shoulder angular velocity variability revealed a main effect of difficulty for the vertical z-axis, F(1, 10) = 8.8, p = .01, ${\eta }_{\mathrm{p}}^2$  = .4. Angular velocity variability was greater in the 70- compared to the 60-bpm condition (60-bpm M = 30.1 degrees/second, SD = 1.2; 70-bpm M = 41.3 degrees/second, SD = 38.1).

Figure 3.

Raw kinematics from are shown from a sample participant. The training and retention data were derived from the 60-bpm condition. Time is expressed as a proportion of the performance (%) because the performances were not of identical length. The raw data for the shoulder angles along the vertical z-axis wherein shoulder angles above zero degrees represent abduction (i.e., angling the body towards the midline) is shown in the top panel. Elbow angular velocity along the z-axis (i.e., wrist speed moving up and down) is shown in the middle panel. Finally, wrist angular velocity along the mediolateral x-axis (i.e., wrist speed moving laterally across the marimba) is shown in the bottom panel.

The ANOVA on left elbow velocity variability along the vertical z-axis revealed a main effect of difficulty, F(1, 10) = 8.5, p = .01, ${\eta }_{\mathrm{p}}^2$  = .5. There was greater variation in elbow joint speed in the 70-bpm (M = 83.1 degrees/second; SD = 10.2) compared to the 60-bpm condition (M = 57.2 degrees/second; SD = 4.1). Peak angular velocity along the vertical z-axis revealed a main effect of difficulty, F(1, 10) = 6.8, p = .03, ${\eta }_{\mathrm{p}}^2$  = .3, and a difficulty × time interaction, F(1, 10) = 6.1, p = .03, ${\eta }_{\mathrm{p}}^2$  = .3. Post-hoc tests for the interaction (HSD = 31.1 degrees/second) revealed greater peak angular velocity in the 70-bpm training (M = 318.1 degrees/second, SD = 37.8) vs. the 60-bpm training condition (M = 241.1 degrees/second, SD = 13.1). Interestingly, peak angular velocity increased from training to retention (M = 279.1 degrees/second, SD = 19.7) in the 60-bpm condition and decreased from training to retention (M = 277.8 degrees/second, SD = 16.3) in the 70-bpm condition. Right elbow velocity variability along the vertical z-axis revealed a main effect of difficulty, F(1, 10) = 19.8, p = .001, ${\eta }_{\mathrm{p}}^2$  = .6, and a difficulty × time interaction, F(1, 10) = 5.2, p = .04, ${\eta }_{\mathrm{p}}^2$  = .3. Post-hoc tests for the interaction (HSD = 15.8 degrees/second) revealed greater velocity variability in the 70-bpm training condition (M = 116.7 degrees/second, SD = 11.1) vs. the 60-bpm training condition (M = 79.8 degrees/second, SD = 6.3). Velocity variability was reduced in retention (M = 91.8 degrees/second, SD = 7.8) vs. training for the 70-bpm condition specifically.

Left wrist peak angular velocity variability along the mediolateral x-axis revealed a main effect of difficulty, F(1, 10) = 5.1, p = .04, ${\eta }_{\mathrm{p}}^2$  = .3, and a difficulty × time interaction, F(1, 10) = 5.1, p = .04, ${\eta }_{\mathrm{p}}^2$  = .2. Post-hoc tests for the interaction (HSD = 5.2 degrees/second) revealed greater peak velocity variability in the 70-bpm training (M = 68.2 degrees/second, SD = 9.1) vs. the 60-bpm training condition (M = 54.2 degrees/second, SD = 6.1). In the 60-bpm condition, peak angular velocity variability increased from training to retention (M = 62.1 degrees/second, SD = 6.1), while decreasing in the 70-bpm condition from training to retention (M = 59.2 degrees/second, SD = 5.1). Finally, right wrist peak angular velocity variability along the vertical z-axis revealed a main effect of time, F(1, 10) = 7.4, p = .02, ${\eta }_{\mathrm{p}}^2$  = .4, with peak angular velocity variability being greater in training (M = 27.1 degrees/second; SD = 4.9) than retention (M = 15.2, SD = 2.9). The results for difficulty × time interactions are shown in Figure 4.

Figure 4.

Kinematic effects involving interactions between timepoint and condition. A) Left shoulder orientation along the mediolateral x-axis. B) Left elbow peak angular velocity along the vertical z-axis. C) Right elbow velocity variability along the vertical z-axis. D) Left wrist peak angular velocity variability along the mediolateral x-axis. Note: error bars = standard error of the mean, * = p < .05.

Performance Ratings

The analysis of performance quality yielded a main effect of time, F(1, 10) = 24.6, p < .001, ${\eta }_{\mathrm{p}}^2$  = .7 wherein retention performances (M = 3.9, SD = .5) were greater than training performances (M = 2.7, SD = .6). The analysis of rhythmic accuracy yielded a main effect of time, F(1, 10) = 6.7, p = .03, ${\eta }_{\mathrm{p}}^2$  = .3, wherein retention performances (M = 3.7, SD = .4) were rated as more accurate than training performances (M = 2.9, SD = .8). Performance quality and rhythmic accuracy findings are shown in Figure 5.

Figure 5.

Ratings of the performances as evaluated by experienced judges are shown. Performance quality and rhythmic accuracy ratings are shown on the left and right, respectively. Note: error bars = standard error of the mean, * = p < .05.

Discussion

The present study assessed motor learning in music via eleven trained percussionists learning a novel excerpt over three training sessions. Participants practiced their assigned excerpt in two tempo conditions during the training sessions. One week after the final training session, participants performed the excerpt again during a retention assessment. The results showed reduced left shoulder abduction in retention vs. training performances. The remaining effects were localized within the velocity domain, wherein peak angular velocity variability of the elbows and wrists was uniquely altered in each difficulty condition as learning progressed. The altered speed of the individual joints coincided with an improvement in overall performance quality and rhythmic accuracy in retention vs. training performances. These findings were interpreted below with respect to mechanisms of motor learning in percussion, and how such knowledge can be applied towards enhancing percussion pedagogy.

The effects of performance difficulty and timepoint were observed predominantly as changes in left limb movements, which was the non-dominant limb for all participants. Interestingly, the left shoulder was angled toward the body's midline to a greater extent during retention compared to training in the 60-bpm condition. This finding is noteworthy given that current percussion pedagogy focuses primarily on movements of the elbows and wrists during training, with limited attention to the role of shoulder movements (see Hartenberger, 2016; Stevens, 2000). There are two plausible technical explanations for the observed difference in left shoulder positioning: the choice of playing spot (i.e., edge vs. middle of the bar) and the performer's interval control (i.e., the distance maintained between mallets in the same hand). For instance, when playing accidentals, performers must often decide between striking the edge of the bar to minimize movement across the keyboard or aiming for the center to achieve a fuller tone. Similarly, players may either keep a narrow interval while compensating with additional arm movements or maintain a larger interval to keep both mallets closer to each target. These choices could impact the degree of shoulder abduction observed, especially during large leaps in the low register (such as the low D♭). It is important to note that the length and width of a marimba bar increases gradually as pitch descends, which means that playing a given interval in the lower register requires a larger distance between the mallets in comparison to the same interval played in the higher register. While these aspects were not directly measured in the current study, they may help explain the increased efficiency of left shoulder positioning during retention. Future research should more systematically examine how control of mallet intervals and strike location affect joint movements involved in gross motor control (i.e., shoulders). Given that post-training performances were associated with greater performance quality and rhythmic accuracy, the present findings highlight the importance of incorporating non-dominant shoulder training into percussion pedagogy. Shoulder orientation likely also contributed to the angular velocity of the elbows and wrists.

Controlling movements of the elbows is critical in music performance. In fine motor control contexts like piano, previous studies have shown angular velocity did not differ between fast and slow tempi (Furuya & Kinoshita, 2008). However, gross motor control contexts like marimba have shown that tempi increased elbow velocity variability, which reduced optimal sound production (Loria et al., 2022b). In the present study, peak angular velocity increased from training to retention timepoints in the 60-bpm condition but decreased over the same timepoints in the 70-bpm condition. Interestingly, the direction of change across the tempi conditions served to bring peak angular velocity to nearly identical speeds at retention. Such findings may be in line with Motor Schema Theory wherein learners first acquire a generalized motor program that defines the structure of an action (e.g., using a certain limb to play a note), followed by the development of the schemata which is a motor execution rule refined via practice and experience (e.g., move the limb at a specific speed to play this note; see: Schmidt, 1975; Schmidt, 2003). The peak angular velocity findings may reflect the development of a general motor schema (i.e., developed during training) that specifies certain joint velocities that contribute to successful performance at retention (for similar findings concerning joint velocity across tempi in drum kit performance see Dahl, 2004, 2011). Such motor schemas may then be applied irrespective of performance tempo to support the learning of new pieces of music. The potential link to Motor Schema Theory may also explain reductions in velocity variability in the right elbow and wrist.

Even the most well-rehearsed movements can be variable. From a motor control perspective, it is well established that optimal performance is associated with a reduction in movement variability (e.g., Harris & Wolpert, 1998; Todorov & Jordan, 2002; van Beers et al., 2004). The results of the present study showed that velocity variability of the right elbow was lower in retention vs. training for the 70-bpm condition, and that right wrist peak angular velocity variability was lower in retention vs. training, irrespective of tempo. The findings of reduced variability of elbow speed in relation to training agree with previous reports showing improved elbow velocity control as a function of learning (i.e., Loria et al., 2022a). However, a novel finding of this study is the co-occurrence between reductions in velocity variability and the increase in overall performance quality and rhythmic accuracy, as indicated by the judges’ ratings. The results of the present study could suggest that reducing variability in elbow and wrist speed may mechanistically improve the performer's ability to produce sound. Such findings may be relevant for developing specific instructional techniques targeting elbow and wrist velocity, which may support the training of future percussionists.

Limitations

This experiment evaluated the relationship between kinematics and performance outcome in experienced percussionists. Because a highly trained sample was used, findings such as the difference in the position of the left shoulder may not be generalizable to beginner percussionists. Furthermore, the task performed by the left and right arms in marimba performance is inherently asymmetric. Although the excerpt used in this study includes a continuous stream of 16th notes and some mirror-like movements between limbs, these movements are not truly symmetrical, particularly along the y-axis. In addition, the larger spacing between bars in the lower register and the associated longer reach required can yield asymmetrical body position in the low register compared to the high register. Future research may address this limitation by designing excerpts involving more symmetrical intervals and isolated movements of each arm to serve as a baseline for comparing left and right limb performance.

This study also did not collect any data specific to participants’ four-mallet technique. It was generally observed, but not recorded, that some participants opted to play all accidentals on the edge of the bar, while others attempted to reach for the middle of the bar. Participants were also not instructed to employ either a Burton, Stevens, or Traditional grip, which may have led to divergent technical solutions to perform note changes (e.g., controlling the distance between the mallets vs. relying on fast arm movements). We also did not provide explicit instructions to avoid the use of mental practice to rehearse the excerpt between the final training and retention sessions. This can be improved upon in future work by providing such instructions. Due to our focus on mirroring typical practice conditions wherein tempi are increased over time as skills increase, we did not counterbalance performances across the tempi conditions. Future research may elect to counterbalance tempi conditions to gain further insight into whether the order at which various tempi are practiced impact the rate or effectiveness of learning. Finally, future work using larger sample sizes may elect to run regression analyses linking movement changes to improved performance outcomes to further strengthen the initial observations provided in the present study.

Conclusion

The present study examined motor learning in marimba performance. The results reveled altered shoulder orientation and lower velocity variability in retention vs. training performances. These changes in individual joint motions coincided with improved ratings of overall performance quality and rhythmic accuracy in retention vs. training performances. The findings highlight that as motor learning progressed, changes in individual joint motions likely improved the performer's ability to produce sound. The results may inform the refinement of training paradigms focused specifically on training shoulder orientation and elbow and wrist speed. Doing so may enhance motor learning outcomes within percussion pedagogy.