Changes in Parkinsonian gait kinematics with self-generated and externally-generated cues: A comparison of responders and non-responders

Abstract

Purpose

Rhythmic auditory stimulation such as listening to music can alleviate gait bradykinesia in people with Parkinson disease (PD) by increasing spatiotemporal gait features. However, evidence about what specific kinematic alterations lead to these improvements is limited, and differences in responsiveness to cueing likely affect individual motor strategies. Self-generated cueing techniques, such as singing or mental singing, provide similar benefits but no evidence exists about how these techniques affect lower limb joint movement. In this study, we assessed immediate effects of external and self-generated cueing on lower limb movement trajectories during gait.

Methods

Using 3D motion capture, we assessed sagittal plane joint angles at the hip, knee, and ankle across 35 participants with PD, divided into responders (n=23) and non-responders (n=12) based on a clinically meaningful change in gait speed. Joint motion was assessed as overall range of motion as well as at two key time points during the gait cycle: initial contact and toe-off.

Results

Responders used both cue types to increase gait speed and induce increases in overall joint ROM at the hip while only self-generated cues also increased ROM at the ankle. Increased joint excursions for responders were also evident at initial contact and toe off.

Conclusions

Our results indicate that self-generated rhythmic cues can induce similar increases in joint excursions as externally-generated cues and that some people may respond more positively than others. These results provide important insight into how self-generated cueing techniques may be tailored to meet the varied individual needs of people with PD.

1. Introduction

Overall reductions in movement amplitude and speed contribute to gait impairment among people living with Parkinson disease (PD). Investigation of the underlying kinematic alterations that lead to the shuffling gait pattern associated with PD reveals key impairments throughout the gait cycle (Sofuwa et al. 2005; Švehlík et al. 2009). These include reduced ankle dorsiflexion at initial contact (IC), related to less effective heel strikes, and reduced hip extension and ankle plantarflexion at toe-off (TO), contributing to smaller steps and less forceful push-off (Kimmeskamp and Hennig 2001; Morris et al. 2005; Mariani et al. 2013). Combined, these kinematic alterations reduce stability and increase fall risk (Morris et al. 2005; Mariani et al. 2013).

Rhythmic auditory stimulation (RAS) is a form of external auditory cueing consistently shown to increase speed and stride length(Ghai et al. 2018), but evidence detailing the effects of cues on lower limb movement trajectories is limited to very few studies (Picelli et al. 2010; Pau et al. 2016; Erra et al. 2019). Picelli et al. assessed cued walking at different cadences and showed improved motor strategies with cues at 110% of preferred cadence indicating that participants adopted a more effective hip flexor motor strategy that allowed them to increase stride lengths(Picelli et al. 2010). Pau et al. found that RAS combined with gait training allowed people with PD to regularize movement by increasing hip and knee ROM but not ankle(Pau et al. 2016). Erra et al. showed similar improvements for participants both ON and OFF meds and that patients exhibiting higher disease severity improved more(Erra et al. 2019). Rinaldi et al. also explored the effects of medication and showed that cues combined with dopaminergic-replacement therapy allowed for increased muscle activation during the initial contact with the floor and the terminal swing phase(Rinaldi et al. 2014). The aforementioned studies are limited to the use of metronome beats, in spite of evidence that musical cues better support entrainment(Rose et al. 2019) and are more motivating for gait therapy (Rodger and Craig 2016). Only one study, by Thaut et al., used musical cues combined with metronome to induce improvements in ankle dorsiflexion after training, however no other joints were assessed(Thaut et al. 2018). Therefore, a more comprehensive exploration of how musical auditory cues affect lower limb joint excursions is needed.

Furthermore, in spite of its proven effectiveness, RAS has notable limitations. First, synchronizing to external sources may negatively affect stride-to-stride variability, a known marker of gait stability (Hausdorff et al. 2007; Brodie et al. 2013; Cochen De Cock et al. 2018). Second, the effects of RAS decline immediately after cue withdrawal, so constant cueing is necessary (Thaut et al. 2018). Third, responsiveness to external cues is highly variable between participants, reducing its effectiveness for some individuals (Dalla Bella et al. 2017). Differences in disease severity(Erra et al. 2019), age(Ghai et al. 2018), and rhythmic ability(Leow et al. 2014) may all play a role in how likely an individual is to respond positively to auditory cues. These limitations highlight a need to find alternatives that are translatable to daily life for a variety of people with PD. Self-generated cueing in the form of singing or mental singing may provide one such alternative. Recent research from our lab suggests that self-generated cues convey similar benefits as external cues without increasing gait variability (Harrison et al. 2017; Harrison et al. 2018). However, no studies to our knowledge have investigated the effects of self-generated cueing on joint kinematics to determine the specific impact on movement patterns. Furthermore, little is known about who may respond best to this technique.

The purpose of this study was to compare gait kinematic profiles during different rhythmic cueing techniques for people with PD. We analyzed sagittal plane movement during gait while using externally-generated and self-generated cues. We hypothesized that both cue types would elicit increases in lower extremity joint movement, and that this effect would be greater for some participants than others.

2. Methods

2.1: Participants

All participants had diagnosed, idiopathic PD. Demographic information is displayed in Table 1. Inclusion criteria were: a) able to stand independently for at least 30 minutes; b) normal peripheral neurological function; c) no history of vestibular disease; and d) no evidence of dementia (Mini Mental State Examination (MMSE) ≥ 24). Participants were excluded if they had any of the following: a) any serious medical problem aside from PD; b) deep brain stimulation surgery; c) diagnosis of peripheral neuropathy; or d) use of dopamine-blocking medication. All participants were recruited as part of a larger study, and only those with a body mass index (BMI) < 30 who were naïve to the cued conditions were included in the present analysis. Of 56 participants in the larger study, 35 met all inclusion criteria.

Table 1.

Participant demographics

	All	Non-responder n=35	Responder n=12	p
Gender (% female)	40	25	47.8	0.282
Age, years	67.17 (9.04)	71.33 (10.47)	65.00 (7.57)	0.048
Years since diagnosis	4.92 (4.82)	3.08 (1.82)	5.88 (5.62)	0.52
Hoehn & Yahr, median (range)χ	2 (1,3)	2 (2,3)	2 (1,2)	0.083
MDS-UPDRS-III	28.69 (11.21)	32.33 (10.45)	26.78 (11.35)	0.168
MMSE, median (range)	29 (25,30)	30 (27,30)	29 (25,30)	--
Uncued velocity, m/s	1.15 (0.18)	1.15 (.20)	1.15 (.18)	0.974
Uncued cadence, steps/min	110.05 (7.81)	105.36 (6.83)	112.51 (7.26)	0.008
Uncued stride length, m/s	1.26 (.19)	1.31 (.21)	1.23 (.17)	0.204
Musical experience, years	3.29 (5.16)	1.33 (2.96)	4.30 (5.80)	0.107
BQMI, median (range)	4 (2,7)	4 (2,7)	4 (2,7)	--
NFOG-Q, n (range)χ	11 (4–26)	2 (6–26)	9 (4–21)	0.522
Fall Status, fallers, nonfallersχ	11, 24	2, 10	9, 14	0.372

Values represent mean ± SD, except where noted. Fallers defined as participants who experienced 1 or more falls in the past 12 months. T-tests and Chi-square tests used for between-group comparisons when appropriate. Bold values indicate p<.05.

Abbreviations: MDS-UPDRS-III, Movement Disorder Society Unified Parkinson Disease Rating Scale – Subscale III. MMSE, Mini Mental Status Examination. BQMI, Betts’ Questionnaire Upon Mental Imagery. NFOG-Q, New Freezing of Gait Questionnaire.

^χChi-square test.

Participants were recruited through the Movement Disorders Clinic at Washington University School of Medicine in St. Louis, Missouri, and through the local chapter of the American Parkinson Disease Association. The study was approved by the Human Research Protection Office at Washington University School of Medicine, and all participants provided written informed consent prior to data collection and were compensated for their time.

2.2: Protocol

Participants were tested in the ‘ON’ state of their anti-Parkinson medication in order to increase relevance of assessment conditions to daily walking. All participants completed a series of evaluations prior to kinematic assessment. Questionnaires included the New Freezing of Gait Questionnaire (NFOG-Q), Fall History Questionnaire, auditory portion of the Betts’ Questionnaire Upon Mental Imagery (BQMI) (Sheehan 1967), and questions about prior musical experience. Disease severity was assessed using the Movement Disorder Society Unified Parkinson Disease Rating Scale – Motor Subscale III (MDS-UPDRS III) (Goetz et al. 2008). Three initial baseline walking trials, measured on a 5m instrumented walkway (GAITRite, CIR Systems, NJ), were used to assess self-selected walking cadence, which was used to tailor individual cue tempos for the kinematic assessments.

2.3: 3D Motion Capture

Sagittal-plane kinematic data were collected using an 8-camera Hawk Digital RealTime system by Motion Analysis (Motion Analysis Corporation, Santa Rosa, CA) with a 100Hz sampling rate. Participants were provided form-fitting clothing and wore their own shoes. Forty-nine reflective markers (20mm diameter) were placed on bony prominences of the lower extremities and pelvis, including a marker placed 0.025m above the floor on the heel of each shoe. The thigh and shank were tracked using plates with four evenly-spaced markers mounted 0.089m above the lateral condyle of the femur and 0.152m above the lateral malleolus of the fibula.

2.4. Procedure

For each movement condition, participants walked diagonally across a 3.05m x 3.05m x 3.05m capture volume. The cue was administered from a laptop connected to speakers no farther than 3.05m from the participant during walking to ensure audibility. The song “Row, row, row your boat” was used for cueing to ensure participants’ familiarity with the melody and lyrics. This particular instrumental version was designed with an easily detectable, salient beat (Harrison et al. 2018). The song tempo was adjusted to 110% of each participant’s self-selected walking cadence, derived from the baseline GAITRite trials, while maintaining key consistency using Audacity open source audio editing software (The Audacity Team, audacity.sourceforge.net/).

Participants walked across the capture volume in four different conditions. Five trials of each condition were collected, and a trained research team member walked behind participants to prevent falls. Participants first walked at their own comfortable pace in silence for the UNCUED condition. Three randomized cued conditions then followed:

1. MUSIC: Participants walked to the beat of the musical cue. Participants listened to the song one time through and began walking as the song played a second time. The music continued playing during walking.

2. SING: Participants listened to the musical cue one time and then began walking while singing out loud at the tempo they just heard. The music was not playing during walking.

3. MENTAL: Participants listened to the musical cue one time and then began walking in silence while singing in their heads. They were not allowed to move their lips or produce overt sound. The music was not playing during walking.

2.5: Data processing

Motion capture data were pre-processed in Cortex (version 1.1.4, Motional Analysis Corporation, CA) and imported into Visual3D (version 6, C-Motion, MD). An initial static trial was collected to create a link segment model for each individual using a modified Helen Hayes pelvis. Joint angles were calculated by the relative rotation of the distal to proximal segments (Wu et al. 2005). Three gait trials in each condition were processed for analysis. A low-pass Butterworth 6Hz filter was used to smooth the kinematic data, and hip, knee, and ankle joint angles and spatiotemporal measures were extracted. Gait cycles were defined by heel strikes, which were calculated by the velocity of the toe marker in relation to the pelvis using a previously validated method (O’Connor et al. 2007). No significant differences were noted between sides, so left and right gait cycles were combined and a minimum of six cycles was used for each participant within each condition. Joint trajectories were normalized to percent of gait cycle, and range of motion (ROM) for each joint was calculated as the difference between maximum and minimum values during the gait cycle.

2.6: Statistics

Participants were divided based on whether or not they demonstrated a clinically meaningful difference in gait speed of 0.06m/s (Hass et al. 2014), a metric that has previously been used to qualify PD responsiveness to auditory cueing (Dalla Bella et al. 2017). We divided our sample based on differences in gait speed between the UNCUED and MENTAL conditions, as we expected to see the greatest improvement in the MENTAL condition based on our previous studies (Harrison et al. 2018; Harrison et al. 2019). This elicited two groups: non-responders (n=12) and responders (n=23). Differences in demographic variables were examined using paired t-tests for continuous variables (or for non-parametric distributions, Mann-Whitney) and Pearson Chi-Square for categorical variables. Statistical significance was set at α≤.05. For demographic information presented in Table 1, participants were also classified as “fallers” if they self-reported one or more falls in the six months prior to testing and as “freezers” if they answered “yes” to the first question of the NFOG-Q, “Did you experience freezing episodes in the past month?”

IBM SPSS (version 24, IBM, NY) was used for all analyses. Differences between conditions and groups were analyzed via repeated measure multivariate analysis of variance (RM-MANOVA) models. Main effects of condition (UNCUED, MUSIC, SING, MENTAL), group (responder, non-responder), and group by condition interaction were analyzed using three separate multivariate models (joint range of motion, angle at initial contact (IC), and angle at toe-off (TO)). Each model assessed sagittal-plane joint kinematics at the hip, knee, and ankle. Mauchly’s test assessed sphericity and Greenhouse-Geisser corrections were used when necessary. Tukey-corrected post-hoc pairwise comparisons were used to examine significant univariate results. Extreme outliers (≥3 interquartile ranges from mean) were winsorized (Dixon 1960).

3. Results

Figure 1 displays joint trajectories normalized to the gait cycle for non-responders and responders for the hip, knee, and ankle in each condition. Our previous papers (Harrison et al. 2017; Harrison et al. 2018; Harrison et al. 2019) focused on spatiotemporal features of gait such as velocity, cadence, stride length, and variability measures, so in this paper we are focusing our results on joint trajectories. Spatiotemporal gait features are provided in Supplemental Figure 1 as a reference.

Figure 1.

Joint angle trajectories for the hip, knee, and ankle show data normalized to the gait cycle and averaged across 12 non-responders (left column) and 23 responders (right column). Fine dotted lines represent ± 1 standard deviation. Thick dotted lines represent joint excursions during UNCUED walking. Non-responders show slight reductions in overall joint range of motion while responders show more expansion in joint range of motion. Differences between non-responders and responders are evident at initial contact (IC, 0% of gait cycle) and at toe off (TO, demarcated by the vertical line).

3.1: Joint range of motion (ROM)

Multivariate tests for ROM indicated no effect of responder group (p=.132) but a significant main effect of condition (F(3,31)=2.02, p<.001, partial η²=.16). Univariate tests indicated a significant main effect of condition for hip ROM only (F(3,99)=16.49, (p<.001) partial η²=.33) with pairwise comparisons showing that ROM at the hip was greater during all cued conditions relative to UNCUED. There was a significant interaction between group and condition at the multivariate level (F(9,25)=3.52, p=.006, partial η²=.66), with concomitant interactions at the univariate level of the hip (F(3,99)=9.46, p<.001, partial η²=.22), knee (F(2.1, 69.7)=3.29, p=.041, partial η²=.09), and ankle (F(3,99)=2.83, p=.043, partial η²=.08). Pairwise comparisons indicate that responders increased ROM at all three joints during cueing as compared to UNCUED (all p<.001) (Fig. 1D–F). Angle-angle plots in Figure 2 illustrate overall reduction in range of movement for non-responders as compared to overall expansion of range of movement for responders, indicating more effective use of cueing.

Figure 2.

Angle-angle plots representing mean joint trajectories for the hip, knee, and ankle plotted against one another. Data are normalized to the gait cycle and averaged across 12 non-responders (left column) and 23 responders (right column). Dotted lines represent joint excursions during UNCUED walking and solid lines represent joint excursions during cued conditions. Non-responders exhibit reductions in movement trajectories during all cued conditions whereas responders show expansion of overall movement during cueing. Abbreviations: IC, initial contact. TO, toe off. Arrows indicate direction of movement along gait cycle.

3.2: Joint angles at initial contact (IC)

Multivariate tests for joint angles at IC revealed a trend toward a significant effect of group (F(3,31)= 2.85, p=.054, partial η²=.216) and a significant main effect of condition (F(9,25)=6.10, p<.001, partial η²=.687). Univariate tests indicated a significant difference of joint angle at IC between conditions at the hip (p<.001), knee (p=.005), and ankle (p=.001). Pairwise comparisons revealed increased joint angles at IC from UNCUED at the hip during all cued conditions (all p<.002) and at the knee and ankle during music and mental (all p<.048). The interaction between group and condition at the multivariate level was significant (F(9,25)=4.04, p=.003) partial η²=.59) with univariate tests showing a significant interaction for the hip (F(3,99)=8.22, p<.001, partial η²=.20), but not at the knee or ankle. Pairwise comparisons indicated that responders had a greater degree of hip flexion at IC in all cued conditions relative to UNCUED (all p<.001) whereas non-responders did not (Fig 1A,D). Quantitative kinematic differences between gait conditions by group are displayed in Table 2.

Table 2.

Gait variables.

Condition		UNCUED	MUSIC	SING	MENTAL
ROM (°)
Hip	Non-responder	37.62(6.16	37.88(6.34)	37.10(6.84)	37.58(6.48)
	Responder	37.10(5.67)	39.86(6.05)*	39.64(6.06)*	40.32(6.56)*
Knee	Non-responder	63.51(4.22)	62.71(5.83)	62.37(3.92)	62.73(4.02)
	Responder	62.29(6.56)	62.58(5.82)	63.22(5.97)	63.17(5.54)
Ankle	Non-responder	27.66(5.17)	27.26(5.37)	27.20(5.43)	27.63(5.21)
	Responder	28.11(4.53)	28.90(4.83)	28.73(4.74)	29.29(4.79)*
Initial contact (°)
Hip	Non-responder	28.26(8.17)	28.61(8.41)	28.68(8.34)	28.29(8.18)
	Responder	29.65(6.51)	31.52(7.02)*	31.03(6.82)*	31.50(7.15)*
Knee	Non-responder	7.76(5.59)	10.53(7.14)*	10.36(7.88)*	10.20(6.91)*
	Responder	8.76(5.29)	9.58(4.55)	9.23(5.45)	9.47(5.17)
Ankle	Non-responder	7.90(4.16)	8.17(4.42)	7.79(4.21)	7.94(4.03)
	Responder	7.73(2.58)	8.70(2.53)*	8.41(2.55)	8.75(2.71)*
Toe off (°)
Hip	Non-responder	3.91(11.68)	5.38(12.26)	4.74(11.99)	4.62(8.60)
	Responder	6.97(6.61)	6.98(6.29)	6.75(6.57)	6.70(6.64)
Knee	Non-responder	53.74(9.03)	54.37(9.51)	54.07(8.95)	53.10(7.43)
	Responder	54.39(6.00)	57.02(6.21)*	56.31(5.94)*	56.79(9.53)*
Ankle	Non-responder	−0.35(7.48)	0.91(9.22)	0.29(7.58)	1.31(8.03)
	Responder	−0.88(6.66)	−4.06(7.56)*	−3.67(7.62)*	−4.1(6.88)*

Values represent mean ± SD across all participants.

^*indicates significant difference from UNCUED according to pairwise comparisons.

Abbreviations: ROM, range of motion.

3.2: Joint angles at toe off (TO)

Multivariate tests for joint angles at TO revealed a significant main effect of condition (F(9,25)=5.54, p<.001, partial η²=.667) but no main effect of group. For the main effect of condition, univariate tests indicate a significant difference of joint angle at TO between conditions at the knee (F(3,99)=20.03, p<.001, partial η²=.38) and ankle ((F(2.3,89.79)=6.49, p<.002, partial η²=.16). Pairwise comparisons revealed greater knee flexion at TO during all cued conditions relative to UNCUED (all p<.001) and greater ankle plantarflexion at TO during sing (p=.017) and mental (p=.006) relative to UNCUED. The interaction between group and condition at the multivariate level was significant (F(9,25)=2.58, p=.029) partial η²=.48) with univariate tests showing a significant interaction at the knee (p=.046) and ankle (p<.001) with a trend at the hip (p=.053). Pairwise comparisons reveal that non-responders had greater knee flexion at TO and less ankle plantarflexion at TO in all cued conditions relative to UNCUED (all p<.001)(Fig 1B,C).

4. Discussion

This is the first study to our knowledge to compare the effects of externally-generated and self-generated musical cues on lower extremity gait kinematics in people with PD. The results suggest self-generated cues are as effective as externally-generated cues in improving lower limb joint trajectories. In fact, self-generated cues may be slightly more effective than external cues since both cue types increased ROM at the hip, but only self-generated cues increased ROM at the ankle as well. A second aim of our study was to explore differences in responsiveness to cueing based on previous literature. By dividing our sample into non-responders and responders, we observed that responders appeared to use rhythmic cues to benefit overall movement amplitude during gait while non-responders showed a slight reduction in amplitude during rhythmic cueing.

From our sample of 35 participants, 23 were classified as “responders” and 12 were classified as “non-responders”. The groups did not differ in most demographic characteristics; however responders were younger and had significantly higher UNCUED cadence. Responders appeared to achieve increases in velocity with the use of cues by concurrently increasing cadence and step length. Responders also increased overall joint angle amplitude throughout the gait cycle, indicating that they were able to utilize auditory cues to effectively combat amplitude dysregulation, and that this overall increase in joint motion likely contributed to an overall increase in velocity.

For responders, increased lower extremity motion during cueing concurs with previous research showing overall increases in ROM during external auditory cueing at the hip (Picelli et al. 2010; Pau et al. 2016) and ankle (though only with cues given at 90% and 100% of preferred cue rate and not at 110%, which differs from our study since we only tested 110%) (Picelli et al. 2010). Here, we show that self-generated cues can elicit similar improvements in ROM at the hip and ankle.

Responders also exhibited improvements in joint motion at key points throughout the gait cycle. At IC, cues elicited increases in flexion at the hip (in all cued conditions) and ankle (in MUSIC and MENTAL). Increased ankle dorsiflexion at IC may indicate a more effective stepping strategy that could reduce shuffling at the beginning of the gait cycle. As attentional strategies to focus on heel strike can cause an immediate increase in ankle dorsiflexion (Ginis et al. 2017), the musical tasks utilized here may have elicited a more vigorous heel strike by increasing emphasis on the downbeat. Responders also showed increased joint excursion at TO, demarcating the beginning of the swing phase of gait. During cueing, responders increased knee flexion and ankle plantarflexion at TO, which may have helped elongate swing times, increase push-off, and improve forward propulsion, translating as an increase in stride length and velocity (Judge et al. 1996).

Non-responders, in contrast, did not significantly alter gait velocity, cadence, or stride length during cued conditions, nor did cueing elicit overall improvements in joint amplitude. Instead, non-responders exhibited subtle overall reductions in movement that may reflect ineffective use of auditory cues. Several observations may account for this. While both groups exhibited similar UNCUED gait velocity, non-responders began with a lower cadence and did not effectively increase cadence during cued conditions. While the increased cue rate of 110% should have elicited a 10% increase in cadence, non-responders achieved, on average across conditions, only a 3.15% increase (compared to 7.05% for responders). This failure to effectively adapt to the higher cue rate of 110% may indicate it was too quick for these participants, causing them to shorten steps in order to match the song tempo. A past study using metronome cues lends support to this idea, as improvements in ankle ROM at 90% and 100% of preferred cue rate disappeared at 110% (Picelli et al. 2010). Evidence that the cue rate was too fast supports recommendations to individually tailor cue rates in order to optimize stride length (Dalla Bella et al. 2017). Inability to alter joint-level kinematics may also reflect an inflexible neuromuscular system in some participants with PD, indicating that auditory cueing may be an ineffective strategy for gait rehabilitation in this group (Kuhman et al. 2018). Such inflexibility may be age-related as age is known to affect ability to effectively utilize auditory cues (Ghai et al. 2018) and non-responders were significantly older than responders.

Alternatively, non-responders may have utilized cues more effectively had we offered instructions to elongate strides while also walking to the beat of the song. Since we did not correct participants if they marched to the beat, increased knee flexion during cueing, seen only in the non-responder group, suggests a tendency to march by lifting their knees, rather than stepping out to elongate steps.

Inability to match the cue rate may also relate to difficulty in detecting the beat or matching movement to it, which would corroborate previous reports linking sensorimotor skills to the success of RAS in improving gait (Dalla Bella et al. 2017). Non-responders, in general, had less musical experience than the responder group, which may have impacted their ability to synchronize to musical cues. Further evidence for higher beat impairment in non-responders could be their significant increase in stride length variability during the MUSIC condition only, suggesting that the external musical cue created a burdensome challenge to their walking performance. Some degree of musical training or rhythmic skill may thus improve the likelihood of responding to musical cues and may be particularly useful in mental singing as it requires maintaining a beat in silence.

Our results should be considered in light of limitations. In spite of statistically significant differences between groups and conditions, overall improvements in joint motion were small. Our participants had an overall low level of disease severity which may reduce relevance of these results to people with higher levels of gait impairment.

In this study, we showed that people with PD can gain immediate benefit to gait with both externally-generated and self-generated musical cues, but that not everyone responds positively to uniform cues. For responders in the present study, all cues tested improved gait kinematics, as evidenced by increased amplitude of lower limb joint trajectories, while non-responders gleaned no immediate benefit from cues. This suggests a need for personalization of cue tempo and possibly training in order for some individuals to better utilize cues to benefit gait. Future work should explore predictors such as age and previous musical experience to determine who is most likely to gain benefit from self-generated and externally-generated auditory cues.

Supplementary Material

Supp 1

Supplemental Figure 1. Spatiotemporal gait parameters and variabilities across participants divided by non-responders and responders. Values represent mean ± SD. * indicates significant difference from UNCUED.