The Effects of Muscle Fatigue and Movement Height on Movement Stability and Variability

Abstract

Performing repetitive manual tasks can lead to muscle fatigue, which may induce changes in motor coordination, movement stability, and kinematic variability. In particular, movements performed at or above shoulder height have been associated with increased shoulder injury risk. The purpose of this study was to determine the effects of repetitive motion induced muscle fatigue on posture and on the variability and stability of upper extremity movements. Ten healthy subjects performed a repetitive task similar to sawing continuously until volitional exhaustion. This task was synchronized with a metronome to control movement timing. Subjects performed the sawing task at shoulder (‘High’) and sternum height (‘Low’) on two different days. Joint angles and muscle activity were recorded continuously. Local and orbital stability of joint angles, kinematic variability (within subject standard deviations), and peak joint angles were calculated for five bins of data spaced evenly across each trial. Subjects fatigued more quickly when movements were performed at the High height. They also altered their kinematic patterns significantly in response to muscle fatigue. These changes were more pronounced when the task was performed at the High height. Subjects also exhibited increased kinematic variability of their movements post-fatigue. Increases in variability and altered coordination did not lead to greater instability, however. Shoulder movements were more locally stable when the task was performed at the High height. Conversely, shoulder and elbow movements were more orbitally unstable for the High condition. Thus, people adapt their movement strategies in multi-joint redundant tasks and maintain stability in doing so.

1. INTRODUCTION

Performing manual tasks repeatedly can lead to muscle fatigue, which may induce changes in motor coordination (Viitasalo et al. 1993; Bonnard et al. 1994; Forestier and Nougier 1998). Some changes in coordination may pre-dispose people to develop injuries by inducing poor biomechanics (Sparto et al. 1997b; Mizrahi et al. 2000). Injury risk during fatiguing repetitive tasks may also be mediated by changes in variability and/or stability of the movement (McQuade et al. 1998; Potvin and O’Brien 1998).

Muscle fatigue is defined as a reduction in the force generating capacity of a muscle or muscle group after activity (Bigland-Ritchie and Woods 1984). In low force tasks, fatigue may not result in a decline in the target submaximal force but can cause an increase in perceived effort (Jones and Hunter 1983). For dynamic tasks, where a reduction in force may not be observable, fatigue can be defined as an inability to sustain a target work rate (Fulco 1995). Research performed on isometric contractions, shows that as fatigue progresses, muscle fiber conduction velocity decreases (Farina et al. 2002), the number of active motor units decreases, they fire more slowly (Bigland-Ritchie and Woods 1984), and become more synchronized (Arihara and Sakamoto 1999). These effects can be seen as a compression of the power spectrum toward lower frequencies using surface electromyography (EMG) (Bigland-Ritchie and Woods 1984). This power spectral compression occurs robustly for dynamic as well as isometric contractions (Potvin and Bent 1997; Gerdle et al. 2000; MacIsaac et al. 2001) and has been shown to correlate well with decreases in force generating capacity (Gates and Dingwell 2010).

Muscle fatigue can lead to increased muscle force variability proportional to the force level (Missenard et al., 2008a). This variability in muscle force output may in turn lead to increased kinematic and kinetic variability (Parnianpour et al., 1988, Selen et al., 2007). Muscle fatigue may also reduce co-activation (Missenard et al., 2008b), which can lead to increased kinematic variability (Selen et al., 2005, Missenard et al., 2008b). Osu et al. tested the hypothesis that the increased muscle activation associated with increased cocontraction should lead to increased signal-dependent noise (Harris and Wolpert, 1998) and therefore to increased variability. However, their results demonstrated that while increased cocontraction did cause increased EMG and torque variability, it actually led to decreased kinematic variability of the end-point in targeted reaching movements (Osu et al., 2004). In multi-joint dynamic tasks, people may alter their biomechanical coordination strategies (Sparto et al., 1997, Cote et al., 2002) or muscle activation patterns (Corcos et al., 2002, Gorelick et al., 2003). These adjustments may serve to minimize changes in overall kinematic variability.

Measures of variability (ie. standard deviations, coefficients of variation, etc.), do not quantify the sensitivity of the neuromuscular control system to perturbations, i.e., ‘stability’ (Dingwell et al. 2000). Local stability can be defined as the ability of the body to correct for small perturbations quickly, without tissue damage (Granata et al. 2004). This is actively controlled by muscle recruitment, muscle stiffness, and reflex responses (Granata et al. 2004). Since muscle function is essential to stability, any factors affecting muscle responses could alter local stability. Muscle fatigue causes decreased proprioception (Myers et al. 1999), increased muscle response times (Wilder et al. 1996; Wojtys et al. 1996), increased muscle compliance (Huang et al. 2006), altered reflexes (Freund 1983; Wojtys et al. 1996), and changes in muscle activation patterns (Bonnard et al. 1994; Corcos et al. 2002). Any of these factors could lead to functional instabilities, possibly increasing the risk of an injury (McQuade et al. 1998). For example, repeated trunk flexion movements became more locally unstable after specific fatigue of the trunk extensors (Granata and Gottipati 2008). In contrast, fatigue induced by prolonged walking caused subjects to slow down and their trunk movements during walking became more locally stable (Yoshino et al. 2004).

Changes in movement patterns with fatigue are well documented for a variety of different tasks. Fatigue alters muscle activation patterns and degrades performance in single-joint arm movements (Jaric et al. 1999; Corcos et al. 2002). Fatigue leads to changes in both muscle firing patterns and joint coordination in sub-maximal jumping and hopping (Viitasalo et al. 1993; Bonnard et al. 1994). Muscle fatigue also led to a change in inter-joint coordination patterns in a whole-body sawing task (Cote et al. 2002). Fatigue of various shoulder muscles has been shown to alter the normal scapulo-humeral rhythm exhibited during shoulder flexion movements (McQuade et al. 1995; McQuade et al. 1998; Tsai et al. 2003). Additionally, repetitive motion-induced fatigue was found to alter shoulder posture and body center of mass position during a reaching task at shoulder height (Fuller et al. 2009). All of the studies to date have examined differences between pre- and post-fatigue conditions. The conclusions of such studies may be limited since they do not provide any information about what happens during the fatiguing process. Only one study to date has tracked how underlying physiological changes at the muscle level (e.g., spectral shifts in EMG) resulting from progressing fatigue in a cycling task correlate with changes in movement kinematics (Song et al. 2009). Additionally, no studies have concurrently quantified kinematics, variability and local dynamic stability, so it is unclear whether the documented changes in coordination actually lead to local instability and/or increased variability. Kinematic variability and local stability are also distinct measures which do not correlate well with each other (Dingwell and Marin 2006). As such, it is possible that fatigue may affect variability and stability differently.

The purpose of this study was to determine how people respond to repetitive motion-induced muscle fatigue. Subjects performed a repetitive sawing task at both sternum (‘Low’) and shoulder (‘High’) levels. We hypothesized that 1) Subjects would alter their movement patterns in response to fatigue, and that these changes would be greater when the task was performed at the High height since this was assumed to be more difficult to maintain, 2) Kinematic variability would increase with fatigue, most likely in response to increased muscle force variability, 3) Shoulder movements would become more locally unstable with fatigue, and 4) Movement variability and local instability would be greater for movements performed at the High height.

2. METHODS

2.1. Subjects

Ten healthy right-handed subjects (four female, six male) participated. Their mean (std) age, weight, and height were 27.9 (2.2) yr, 72.4 (18.2) kg and 1.73 (0.10) m, respectively. Prior to the experiment, all participants signed institutionally approved consent forms and were screened to ensure that no subject had a history of medications, surgeries, injuries, or illnesses that might have affected their upper extremity joint movements. To determine handedness, subjects completed a modified version of the Edinburgh Inventory (Oldfield 1971). This inventory indicates the level of dominance of one hand over another. A score of 0/10 indicates a complete left-handed preference, while a score of 10/10 indicates a complete right-handed preference. All subjects scored at least 9/10 on the Edinburgh Inventory, indicating a strong right-handed dominance.

2.2. Experimental Protocol

Subjects performed a repetitive task at two different work heights on two separate visits to the laboratory spaced approximately one week apart. The order of testing was randomized such that the order of presentation was balanced across subjects (Table 1). On each visit, subjects were seated in a device built to simulate a repetitive work task similar to sawing (Fig. 1). During each experiment, subjects made bi-directional horizontal movements in the anterior-posterior direction with their right arm while holding a handle mounted to a metal platform sliding on a low friction track attached to a support frame. Inertial resistance was supplied by an adjustable set of weights mounted on the carriage. Therefore, the resisting load was always opposed to the direction of motion so that the arm extensors were the primary agonists during the pushing stroke, while the flexors were the primary agonists on the pulling stroke. Subjects performed this task continuously until voluntary exhaustion.

Table 1.

Gender, testing condition, and time to fatigue for each subject.

Subject	Gender	First Condition	Time to fatigue (min)
			High	Low
1	F	High	14:11	13:29
2	M	Low	10:44	5:41
3	M	High	29:04	14:22
4	M	Low	30:59	9:56
5	F	Low	15:05	13.02
6	M	Low	25:57	14:49
7	F	High	5:57	5:05
8	M	High	34:14	13:02
9	F	Low	41:12	5:00
10	M	Low	37:53	9:00

Figure 1 –

Subjects were seated in a high-back chair and restrained by belts across the waist and shoulders. A handle with an adjustable weight stack was able to slide with low friction across a horizontal track. This track was adjusted to either the level of the subject’s sternum or their shoulder joint center.

The device was adjusted so that subjects sat with a knee angle of 90°. The height of the metal track was adjusted so the midpoint between the third and fourth finger was either at the level of the xiphoid process (‘Low’) or in line with the shoulder joint center (‘High’). The horizontal position of the chair was adjusted to be comfortable for the subject and to allow for a full range of motion. This was defined as a maximum point almost to full extension (no hyperextension) and a minimum point at the level of the sternum. Subjects also wore a five-point harness (Corbeau, Sandy, UT, USA) across their waist and shoulders, to restrict trunk motion.

To ensure the task resistance was comparable across subjects, we first measured each subject’s maximum pushing/pulling force using a second custom handle attached to a Baseline^® dynamometer that was rigidly mounted on a table. Subjects alternately pushed and then pulled on this rigidly fixed handle with maximal effort three times for five seconds each time with at least one minute of rest in between each attempt. The average of these six peak forces applied during each maximal effort defined that subject’s maximum isometric pushing/pulling strength (MVC). Subjects performed the sawing task with the weight mounted on the carriage (Fig. 1) set to 15% of this maximum pushing/pulling force. The same absolute load was applied in both conditions. The actual forces experienced by each subject were a function of this external load, their hand acceleration (through F = m·a), and to some extent friction. Because movement distance and frequency were also both scaled to each subject, so were these external forces. “This percentage was chosen from pilot testing in the ‘Low’ condition, so that subjects would reach voluntary exhaustion in approximately 15–20 minutes.

Subjects were instructed to synchronize their movements with a metronome. To ensure that the task was dynamically equivalent across subjects, the frequency of the metronome was set to twice the average of the predicted resonant frequencies of the upper arm and forearm segments of each subject (2 beats per each movement cycle) (Gates and Dingwell 2008). The average frequency of movement was 1.07 Hz. To minimize any learning effects, subjects were asked to perform a warm up trial, moving in time with the metronome, for a minimum of 30 seconds (~30 cycles) or until they felt completely comfortable with the task. Pilot testing confirmed that subjects were able to learn this simple task within just a few (< 10) movements (Gates and Dingwell 2008). Subjects then rested for one minute to minimize any fatigue effects that may have occurred during this practice period.

Nineteen reflective markers were placed on the right arm and trunk to define the movements of four body segments. Markers were placed on the trunk at the right and left acromion processes, sternal notch, and seventh cervical vertebra. Clusters of four markers each were placed on the upper and lower arms to define the segments. The hand was defined by four markers at the radial and ulnar epicondyles of the wrist and third and fifth metacarpal-phalangeal joints. Additional markers were placed on the medial and lateral humeral epicondyles for a static calibration trial. A final marker was placed on the top of the handle to define the beginning and end of each cycle. The three-dimensional position of these markers was recorded continuously during all trials at 60 Hz using an eight camera Vicon-612 motion analysis system (Oxford Metrics, Oxford, UK).

Nine pre-amplified electromyography (EMG) surface electrodes (Delsys Inc., Boston, MA, USA) were attached to the dominant arm and torso muscles to record activity in the middle trapezius, pectoralis major, deltoid (anterior, middle and posterior), triceps (lateral head), biceps, flexor carpi radialis, and extensor carpi radialis longus. Electrodes were positioned over each muscle according to accepted recommendations (Konrad 2005). EMG signals were recorded at 1080 Hz using a Delsys Bagnoli-8 (Delsys Inc., Boston, MA, USA) system integrated into the Vicon-612 motion analysis system. Additionally, subjects were asked to rate their perceived exertion (RPE) every three minutes during each trial on a modified Borg scale (Borg 1974; Borg 1982).

2.3. Data Analysis

The instantaneous mean power frequencies (IMNF) of the EMG signals were calculated using wavelet transform methods, as this method is more accurate and robust than the Fourier transform for analyzing nonstationary signals (Hostens 2004). Briefly, a continuous wavelet transform (CWT) of the signal was taken (Matlab 7.0, Mathworks, Natick, MA, USA). The instantaneous mean frequency (IMNF) is calculated by

$$\mathit{\text{IMNF}}=\frac{\underset{ls}{\overset{hs}{\int }}s\times {\left|CWT(s,\tau )\right|}^{2}ds}{\underset{ls}{\overset{hs}{\int }}{\left|CWT(s,\tau )\right|}^{2}ds}$$ (1)

Where ls is the lowest scale of interest and hs is the highest. A ‘daubechies’ wavelet (db5) with a center frequency of 720 Hz at the lowest scale was used for all analyses. This wavelet was scaled in 1-scale intervals from 1–38, which corresponds to center frequencies ranging from approximately 19–720 Hz. IMNF values were averaged over each cycle to give a single IMNF per cycle. These IMNF values quantified how the local fatigue state of each muscle changed across consecutive cycles during each trial (Gates and Dingwell 2010). Localized muscle fatigue would cause the EMG mean frequencies to decrease (DeLuca 1997).

Marker data were filtered using a fifth order Butterworth filter with a cutoff frequency of 15 Hz. Segment coordinate systems were calculated based on the marker positions with a least-squares algorithm (Veldpaus 1988). The joint centers at each instant in time were then calculated based on the position of the joint markers during the static trial (Schmidt et al. 1999). Local coordinate systems were then defined using the International Society of Biomechanics’ (ISB) recommendations for the humerus and forearm (Wu et al. 2005) and a modified coordinate system for the trunk (Hingtgen et al. 2006) and wrist (Rao et al. 1996). The three dimensional movements of the right arm were determined using Euler angles. The rotation sequences used were in accordance with ISB recommendations (Wu et al. 2005). All joints were assumed to have three rotational degrees of freedom. The second rotational angle of the elbow, the carrying angle, is not reported since it changes only minimally due to the biomechanical constraints on the elbow joint.

The maximum angle (‘PeakAng’) within each cycle was identified and used to evaluate changes in overall kinematics. Variability was quantified as MeanSD: the average width of the standard deviation across the movement cycle (Dingwell and Marin 2006). The mean values of IMNF, PeakAng, and MeanSD were calculated by splitting the data into five non-overlapping bins. Only the last 50 cycles in each bin were analyzed to maintain consistency across subjects and conditions.

Local dynamic stability was defined as the quantitative response of the system’s state variables to small perturbations (Kang and Dingwell 2006). To calculate this we first defined a multi-dimensional state space for each joint consisting of its three rotational angles and angular velocities (except at the elbow, which had only two).

$$S\left(t\right)=\left[{\theta }_1\left(t\right),{\theta }_2\left(t\right),{\theta }_3\left(t\right),{\dot{\theta }}_1\left(t\right),{\dot{\theta }}_2\left(t\right),{\dot{\theta }}_3\left(t\right)\right]$$ (2)

These state space descriptions were shown to adequately define the joint motion (Gates and Dingwell 2009). The mean local divergence of nearest neighbor trajectories was calculated using a previously published algorithm (Rosenstein et al. 1993). For any trajectory in state space, that trajectory’s nearest neighbor represents what might happen if a small local perturbation were applied to the system. Short-term local divergence exponents were estimated from the slopes of linear fits to curves using:

$${\lambda }^{*}{}_s=\frac{1}{\Delta t}\langle \text{ln}\left[d_j\left(t\right)\right]\rangle$$ (3)

where $\langle \text{ln}\left[d_j\left(t\right)\right]\rangle$ represents the mean logarithm of the divergence, for all pairs of nearest neighbors, j, throughout a time span (Rosenstein et al. 1993). Positive exponents indicate local instability (i.e., small perturbations grow larger with time), and larger exponents indicate greater sensitivity to local perturbations.

Since the intrinsic time scales were different for each subject (i.e. different average cycle times), the time axes of these curves were re-scaled by multiplying by the average cycle frequency for each subject during each condition. Short-term exponents (${\lambda }_s^{*}$) were calculated from the slopes of linear fits to the divergence curve between 0 and 1 cycle. Here, ‘0’, represents the beginning of each initial perturbation, which could occur anywhere within the 50 cycle block. Long-term exponents (${\lambda }_L^{*}$) were calculated as the slope between 4 and 10 cycles (Dingwell et al. 2007). The short-term exponents represent the initial response to very small perturbations, where the greatest divergence of initially neighboring trajectories occurs. The long-term exponents represent the longer term affect of these perturbations. In this case, the range of 4–10 cycles was chosen because it appeared to be highly linear (Dingwell et al. 2000). Since these divergence curves are sensitive to the number of cycles (Bruijn et al. 2009), only data from the last 50 cycles of each bin were used for consistency across subjects and conditions. This method is also sensitive to the number of points (Granata and England 2006), so the data in each bin were resampled such that each series of 50 cycles had exactly 5000 points.”

Second, orbital stability was quantified by calculating maximum Floquet Multipliers (FM) based on established techniques (Hurmuzlu et al. 1996; Donelan et al. 2004). The state space data for each individual cycle were time-normalized to 101 samples (0% to 100%). We defined a Poincaré section for each percent of the movement cycle, which defined 101 Poincaré sections for the system. 0% and 100% were minimum positions of the handle marker while 50% was the maximum position. The state space, S_k, for each cycle at that Poincaré section evolved to the state at the following cycle, S_k+1 according to the Poincaré map:

$$S_{k+1}=F\left(S_k\right)$$ (4)

The average trajectory across all cycles within a trial was chosen to represent the ‘limit cycle’ for the movement. Limit cycles correspond to single fixed points in each Poincaré plane (i.e., at each % of the cycle). Orbital stability at each Poincaré section was estimated by computing how quickly small perturbations away from these fixed points grow or decay, using a linearized approximation of equation 4:

$$\left[S_{k+1}-S*\right]\approx J\left(S*\right)\left[S_k-S*\right]$$ (5)

where k enumerates individual cycles, S_k is the system state for cycle k at that Poincaré section, and S* is the system state at the fixed point. J(S*) is the system Jacobian matrix for each Poincaré section. The FM are the eigenvalues of J(S*) (Hurmuzlu and Basdogan 1994; Nayfeh and Balachandran 1995). In this case, J(S*) is a 6×6 matrix so there were six eigenvalues. Since each Poincaré section occupied a 5×5 sub-space of S, only the first five eigenvalues were non-zero (Kang and Dingwell, 2008). If these FM had a magnitude < 1, perturbations shrank, on average, by the next cycle, and the system remained stable. In this case, a larger maximum FM indicates that movements are approaching instability (i.e. it takes longer for perturbations to be ‘absorbed’). The magnitudes of the maximum FM for each Poincaré section (i.e., for each % movement cycle) were calculated for all cycles within each bin. We observed small fluctuations in these values, but no discernable trends, consistent with previous results for human walking (Dingwell et al. 2007). We therefore extracted the largest value of MaxFM that was observed across the movement cycle. We would have obtained essentially the same results with almost any other choice of Poincaré section. Data were averaged across subjects to quantify differences between conditions and across bins

Comparisons for peak angles (PeakAng), kinematic variability (MeanSD), dynamic stability (${\lambda }_s^{*}$, ${\lambda }_L^{*}$, and MaxFM) and IMNF were made using 2-factor (Bin (1–5) × Condition (High/Low)) repeated measures ANOVAs (SPSS Inc., Chicago, IL USA). The significance level was set at p < 0.05 for all comparisons. Least significant difference tests were performed to assess significance between bins for each condition and between conditions for each bin.

3. RESULTS

Subjects performed the task for significantly longer in the Low condition (24.1 ± 10.3 min) than for the High condition (12 ± 3.9 min; F = 11.676, p = 0.007). There was no difference in time to fatigue for the female and male subjects (F = 1.015, p = 0.33). All subjects exhibited localized muscle fatigue as measured by decreased IMNF of the EMG signals (Fig. 2). These decreases were statistically significant for all nine muscles tested (p < 0.032). There were no differences in IMNF between the two conditions. However, since the “High” task led to exhaustion much faster than the “Low” task, the IMNF’s would have decreased at a much faster rate in the “High” task when assessed in real time (i.e., minutes & seconds). All subjects reported an RPE of 10 at the completion of each experiment.

Figure 2 –

The average instantaneous mean frequency (IMNF) of the EMG signals were calculated for each of 5 bins. IMNF is measured in inverse scales. This was multiplied by 1000 for visual clarity. Error bars are ±95% confidence intervals. A significant main effect for bin was found for all muscles (p < 0.032). There was no significant effect of condition for any muscle tested. Bins are offset for clarity. ‘*’ represent bins which were significantly different from bin 1 for both conditions.

The average maximum angle (‘PeakAng’) the subjects achieved during each movement varied across the trial (Fig. 3). Subjects tended to lower (increased humeral ‘negative’ elevation (F = 5.604, p = 0.001), and externally rotate (F = 7.934, p = 0.000) their humeri. They also increased their elbow extension (F = 3.827, p = 0.011) and decreased their wrist extension (F = 5.255, p = 0.002) and wrist pronation/supination (F = 4.101, p = 0.008). There were significant High/Low condition effects for humeral plane angle (F = 5.455, p = 0.044), humeral elevation angle (F = 49.692, p = 0.000), elbow flexion angle (F = 16.283, p = 0.003), elbow pronation/supination angle (F = 34.836, p = 0.000), and wrist abduction/adduction angle (F = 14.075, p = 0.005). There were also significant bin × condition interactions for humeral elevation angle (F = 2.716, p = 0.045) and elbow pronation/supination angle (F = 3.689, p = 0.013). In both cases, the changes in peak angle with bin only occurred when movements were performed at the High height.

Figure 3 –

The average maximum angle (‘PeakAng’) for each bin is shown for the High and Low conditions. The top row shows data for the shoulder joint angles (from left to right, humeral plane, elevation angle, internal external rotation). The middle row is elbow flexion/extension and pronation/supination. The bottom row is the wrist flexion/extension, ab/adduction and pronation/supination. ‘^’ represents a significant condition effect, ‘†’ represents bins which are significantly different from bin 1 for the high condition, ‘§’ is for the low condition and ‘*’ is for both conditions. Error bars are ±95% confidence intervals.

The variability (‘MeanSD’) of the movements changed with condition and bin (Fig. 4). The humeral elevation angle was more variable for the High condition (F = 5.683, p = 0.041), while humeral rotation angles were less variable (F = 9.749, p = 0.012). There was a significant bin effect for the humeral plane angle (F = 3.633, p = 0.014), elbow flexion angle (F = 3.246, p = 0.023), wrist flexion angle (F = 4.797, p = 0.003) and wrist abduction/adduction angle (F = 2.845, p = 0.038). These differences were relatively small, however (< 2°).

Figure 4 –

MeanSD for each joint angle is shown across the 5 bins. The top row shows data for the shoulder joint angles (from left to right, humeral plane, elevation angle, internal external rotation). The middle row is elbow flexion/extension and pronation/supination. The bottom row is the wrist flexion/extension, ab/adduction and pronation/supination. ‘^’ represents a significant condition effect, ‘†’ represents bins which are significantly different from bin 1 for the high condition, ‘§’ is for the low condition and ‘*’ is for both conditions. Error bars are ±95% confidence intervals.

Short-term (${\lambda }_s^{*}$) and long-term (${\lambda }_L^{*}$) local divergence exponents (Fig. 5), were calculated to quantify movement stability. Subjects exhibited more locally unstable shoulder movements over the short-term in the Low condition (F = 16.376, p = 0.003). There were no significant main effects for either condition or bin for the long-term local divergence exponents. There were no differences across bins for any variable.

Figure 5 –

The top row shows data for short-term exponents, ${\lambda }_s^{*}$. The bottom row is data for the long-term exponent, ${\lambda }_L^{*}$. Data is shown for the shoulder, elbow and wrist. There was significant main effect for condition (i.e. Low vs. High) at the shoulder for ${\lambda }_s^{*}$ (p = 0.003). There were no significant differences across bins for any variable and no Low/High condition effects for any of the other five comparisons. Error bars are ±95% confidence intervals.

Maximum Floquet multipliers were calculated as a secondary measure of movement stability. Movements were more orbitally unstable at the High height for the shoulder (F = 5.609, p = 0.042) and elbow (F = 5.752, p = 0.040) (Fig. 6). There were no significant differences across bins, however.

Figure 6 –

Orbital Stability Results are shown for the shoulder (SHO), elbow (ELB) and wrist (WRT). MaxFM exhibited shows some differences between conditions (shoulder: p = 0.042, elbow: p = 0.040) where movements were slightly less stable (larger MaxFM) for the High condition. There were no differences across bins. ^ represents a significant main effect for condition (p < 0.05). Error bars are ±95% confidence intervals.

4. DISCUSSION

Muscle fatigue can lead to increased variability (Parnianpour et al. 1988; Selen et al. 2007; Missenard et al. 2008a), and decreased stability (Granata and Gottipati 2008) of repetitive movements. In multi-joint dynamic tasks, people can change their biomechanical coordination strategies (Sparto et al. 1997a; Cote et al. 2002) and/or muscle activation patterns (Corcos et al. 2002; Gorelick et al. 2003). These adjustments could minimize changes in overall kinematic variability and/or movement stability. The purpose of this study was to determine how muscle fatigue affected kinematic variability, local and orbital dynamic stability, and coordination during a repetitive task performed at two work heights. We hypothesized that subjects would exhibit increased variability and decreased stability in response to fatigue. Subjects’ exhibited increased kinematic variability with progressing fatigue, but no changes in movement stability. Subjects also altered their movement patterns in response to fatigue. These kinematic changes were consistent with other studies on fatiguing repetitive movements (Sparto et al. 1997b; Cote et al. 2002), and may have served to minimize changes in movement stability.

The fatiguing protocol effectively fatigued the muscles of the right arm. Each subject reported that their rate of perceived exertion was a 10 (on a scale of 10) when they completed the study. The instantaneous mean frequency of the EMG signals also decreased significantly for both work heights (Fig. 2). Although traditional definitions of muscle fatigue include a ‘decrease in force production’, it was not possible to perform maximum voluntary contractions during this test due to the continuous nature of the task. As such, we were not able to directly quantify decreased force generating capacity of subjects’ muscles using this protocol. However, in a previous study of a very similar task, using the same EMG analysis methodologies, we did show a direct correspondence between the decreases in EMG mean frequencies and decreased force generating capacity (i.e., MVC) (Gates and Dingwell 2010). Subjects’ muscle activity was monitored continuously throughout the task using surface EMG. Surface EMG has been used extensively to quantify changes in muscle function during fatiguing isometric tasks. Dynamic contractions involve changes in muscle length and force that introduce nonstationarities that may sometimes make the interpretation of these measures suspect (Bonato et al. 2001; MacIsaac et al. 2001). That is, EMG spectral content may be distorted by changes in muscle length (Doud and Walsh 1995) and/or by the speed of movement (Masuda et al. 2001). However, a recent modeling study demonstrated that the nonstationarities associated with dynamic contractions have only a negligible impact on mean power frequencies (MacIsaac et al. 2001). These results are supported by several studies which showed that even traditional approaches to estimating MNF and/or MDF can provide valid indicators of muscle fatigue during dynamic contractions (Potvin and Bent 1997; Gerdle et al. 2000; MacIsaac et al. 2001; Nussbaum 2001). The EMG processing procedures adopted in the present study were based upon, and consistent with, the procedures used in these previous studies. Therefore, we are confident that our IMNF results (Fig. 2) validly demonstrate that these muscles had indeed significantly fatigued during these trials.

The High height required the arm to be elevated further against gravity and put the humerus in more impinged position relative to the scapula. This increased the difficulty of the task, as evidenced by the fact that subjects fatigued (i.e., reached volitional exhaustion) approximately twice as fast for the High height task as for the Low height task. Subjects also altered their movement patterns in response to fatigue for both conditions, but these alterations were more pronounced in the High condition (Fig. 5). Specifically, subjects’ humerii became less elevated and more externally rotated with fatigue when movements were performed at the High height (Fig. 3). They also exhibited increased elbow flexion and slightly decreased elbow pronation (Fig. 3). These changes may have served to put the humerus in a less impinged position. Thus this postural adaptation did not increase their risk of shoulder impingement. Similar changes were observed in a study of repetitive reaching at shoulder height while standing. Subjects in that study raised their shoulder joint in order to perform the movement at the same height while decreasing their shoulder abduction angle (Fuller et al. 2009).

Movement stability did not vary across the trial, contrary to our third hypothesis. Additionally, movements at the higher height were more locally stable (Fig. 5) but less orbitally stable (Fig. 6) than those performed at the Low height. In and of themselves, the observed differences between the Low and High height conditions do not reveal why these changes occurred. They could be related to biomechanical changes, or they could be related to changes in neural control. Similarly the lack of significant changes with muscle fatigue may be related to postural adjustments or changes in neural strategy.

One possible reason for the minimal changes in stability with muscle fatigue may have been adjustments in movement speed. Decreasing speed has been shown to increase local stability (Dingwell and Marin 2006; Granata and England 2006). In the present study, we attempted to control for movement speed by having subjects synchronize their movements with a metronome. Subjects were largely able to maintain a constant cycle time across the trial (F = 2.246, p = 0.083) for both conditions (Fig. 7). There were no specific targets for the movement distance, however, so some subjects adjusted their reaching distance across bins (F = 3.313, p = 0.021, bin × condition interaction: F = 5.118, p = 0.002). The average speed therefore also varied across bins (bin: F = 4.122, p = 0.008; bin × condition interaction: F = 6.565, p = 0.000). Using post-hoc analysis to explore these significant interaction effects, we found that temporal parameters were only affected when movements were performed at the High height. Subjects were not able to maintain a constant cycle time as dictated by the metronome. They tended to slow down and make shorter movements over the course of the experiment. Since subjects slowed down, it is possible that this may have offset any effect of fatigue on the stability of movements performed at the High height.

Figure 7 –

Average cycle times, distance and speed across bins are shown for both the Low and High conditions. Error bars represent ±95% confidence intervals about the mean. § indicates bins that were statistically different from bin 1 for the High condition.

Another possible reason for the lack of changes in stability with muscle fatigue may have been changes in muscle cocontraction. To compensate for decreased feedback and delayed responses, the body could increase muscle cocontraction to increase joint stiffness. Some studies have shown such increased cocontraction in response to fatigue (Granata and Orishimo 2001; Grondin and Potvin 2009), while others have found decreased cocontraction (Missenard et al. 2008b; Gates and Dingwell 2010). Previously, we found that subjects performing a similar task did not cocontract and yet maintained/increased the stability of their upper extremity movements (Gates and Dingwell 2010). In addition to stability, cocontraction might also help people maintain movement accuracy (Gribble et al. 2003; Missenard et al. 2008b). In the present study, subjects presumably had to maintain movement stability while also accurately matching the metronome. Clearly these subjects did not place accuracy demands above maintaining stability when performing movements at the High height since they began increasing their cycle time (Figure 7).

Subjects exhibited small, but significant, increases in kinematic variability with muscle fatigue for several joint angles (Fig. 4). However, this kinematic variability also changed differently for different joints. Thus, greater fatigue was not always associated with increased variability. However, we did not observe any variable for which greater fatigue was associated with decreased variability. Likewise, the increase in height of the task did not affect all of the joints studied in the same manner. Most joint angles were unaffected by movement height, while the variability of the humeral elevation angle was greater at the High height and the variability of the humeral rotation angle was greater in the Low condition (Fig. 4). It is possible that the increases in variability were associated directly with increased neuromuscular fatigue. Alternatively, it is also possible that this kinematic variability resulted from adaptations the subjects made to actively combat fatigue. Finally, those changes that were observed, while statistically significant, were modest (< 2 degrees). In this paper, we quantified a large number of parameters (Figs. 3–7). It is likely that not all of these were parameters are independent. As such, some caution is likely warranted in interpreting the degree of statistical significance present in some cases. However, we did observe increased variability in 5 of the 8 joint angles measured, which seems unlikely to have occurred purely by chance.

For most of the biomechanical and dynamic stability variables tested here, obvious trends over time were sometimes difficult to indentify because of the large between-subject variability, particularly for time to exhaustion. It is possible our sample size (n=10) was merely too small to detect significant changes over time for the dynamic stability measures (Figs. 5–6). We did, however, have ample statistical power to detect statistically significant differences between the two tasks for these measures and changes over time for several other kinematic variables (Figs. 3–4, 7). Another possible reason we found no changes in dynamic stability over time for these stability measures might be that by continually altering their kinematics over time, subjects were able to maintain some desired level of stability. Since this was (by design) a redundant task, there were numerous strategies subjects could use to compensate for fatigue. For example, each subject showed a unique pattern as to which muscles were most affected by fatigue during the task. This could have been due to differences in muscle anatomy, muscle fiber type composition, training, etc. All of these factors could potentially affect which muscles fatigued first and to what extent. The differential effects of the task on fatigue led subjects to exhibit different changes over time in their kinematics. The between-subject variability observed in this study was similar to that observed in previous studies of fatigue in complex multi-joint tasks (von Tscharner 2002; Madigan and Pidcoe 2003; Gates and Dingwell 2010). In spite of these large between-subject variations, however, several consistent trends were still observed, particularly for the average kinematic measures (Fig. 3) and kinematic variability measures (Fig. 4).

An additional source of between-subject variability may have come from the mixed genders of the subjects. Previous work has shown that there are gender differences in the effects of fatigue. Specifically, women exhibit increased endurance time than men during sub-maximal, isometric and dynamic tasks (Maughan et al. 1986; Hunter and Enoka 2001) perhaps due to a greater ability to utilize oxidative metabolism than men (Kent-Braun et al. 2002; Russ and Kent-Braun 2003). These gender differences disappear when subjects are matched for strength (Hunter and Enoka 2001), under ischemic conditions (Russ and Kent-Braun 2003), and as the intensity of the exercise increases (Maughan et al. 1986). The subjects in this study performed a low intensity task where the weight applied was titrated to their maximal strength. Thus, the magnitude of the weight applied for the female subjects was less than the male subjects (Female: 9.25 lb, Male: 15 lb, p = 0.013), but there was no difference in endurance time between the female and male subjects (p = 0.33).

In summary, subjects significantly altered their kinematic patterns in response to muscle fatigue. These changes were more pronounced when the task was performed at a higher height. Subjects also exhibited increased variability of their movements post-fatigue. Increases in variability and altered coordination did not lead to changes in local or orbital dynamic stability, however. Local stability of the shoulder was lower when movements were performed at a lower height. In contrast, orbital stability of the shoulder and elbow was lower for movements at the higher height. This research showed that people continuously adapt their strategies in multi-joint redundant tasks and maintain stability in doing so.