Altered activation of the antagonist muscle during practice compromises motor learning in older adults

Abstract

Aging impairs the activation of muscle; however, it remains unclear whether it contributes to deficits in motor learning in older adults. The purpose of this study was to determine whether altered activation of antagonistic muscles in older adults during practice inhibits their ability to transfer a motor task ipsilaterally. Twenty young (25.1 ± 3.9 yr; 10 men, 10 women) and twenty older adults (71.5 ± 4.8 yr; 10 men, 10 women) participated. Half of the subjects practiced 100 trials of a rapid goal-directed task with ankle dorsiflexion and were tested 1 day later with elbow flexion (transfer). The rest did not perform any ankle practice and only performed the task with elbow flexion. The goal-directed task consisted of rapid movement (180 ms) to match a spatiotemporal target. For each limb, we recorded the EMG burst activity of the primary agonist and antagonist muscles. The rate of improvement during task acquisition (practice) was similar for young and older adults (P > 0.3). In contrast, only young adults were able to transfer the task to the upper limb. Specifically, young adults who practiced ankle dorsiflexion exhibited ∼30% (P < 0.05) lower movement error and ∼60% (P < 0.05) lower antagonist EMG burst activity compared with older adults who received equal practice and young adults who did not receive any ankle dorsiflexion practice. These results provide novel evidence that the deficient motor learning in older adults may be related to a differential activation of the antagonist muscle, which compromises their ability to acquire the task during practice.

the ability to learn new motor tasks is essential from infancy to older age. In older age, motor learning becomes of particular importance because of significant neural changes (Eisen et al. 1996; Enoka et al. 2003; Henderson et al. 1980; Narici and Maffulli 2010). Older adults, furthermore, often have to adapt to new motor demands following debilitating neurological diseases that are associated with aging, such as stroke or Parkinson's disease (Akushevich et al. 2013). Although it is well established that the neural activation of muscle changes with aging (Christou et al. 2007; Enoka et al. 2003; Poston et al. 2008a), it is not clear whether these changes contribute to the deficient motor learning in older adults (Osu et al. 2002; Panzer et al. 2011; Seidler 2006). The focus of this study was to determine whether differential activation of the agonist and antagonist muscles during practice compromises motor learning in older adults.

Practice of a motor task is an essential component of motor learning. It allows the learner to extract and acquire important information about the task (Wolpert et al. 2011). With practice, the neural connections that represent the task become relatively permanent (motor learning) and can be used to accomplish similar tasks (generalizability) (Kantak and Winstein 2012). Therefore, motor learning is often tested with transfer tasks. There is evidence that the ability of older adults to transfer motor tasks is compromised relative to that of young adults. For example, after training of sequential movements with the elbow, older adults exhibit an impaired ability to retain the task (Shea et al. 2006) and transfer it to the contralateral arm (Bo et al. 2009; Panzer et al. 2011). Furthermore, Summers and colleagues demonstrated that older adults exhibit an impaired ability to transfer a ballistic movement task from the right to the left index finger (Hinder et al. 2011). To our knowledge, however, no studies have examined ipsilateral transfer in older adults. Ipsilateral transfer will eliminate the interhemispheric interaction that occurs during training and remove the transfer of the neural connections that represent the motor task to the contralateral motor cortex (Hinder et al. 2011; Wang et al. 2011). Therefore, an ipsilateral transfer paradigm may provide a cleaner model for testing motor learning (transfer) than a contralateral transfer paradigm.

The deficient motor learning in older adults may relate to their inability to extract task-relevant information during practice because of impaired movement performance (Wolpert et al. 2011). Impaired movement performance in older adults has been associated with altered activation of the involved muscles (Christou 2011; Enoka et al. 2003). For example, greater variability in the motor unit discharge is associated with greater force (Laidlaw et al. 2000) and movement (Kornatz et al. 2005) variability in older adults. In addition, higher coactivation of antagonistic muscles is related to greater force variability in older adults (Tracy and Enoka 2002). However, during goal-directed contractions, which we use in this study, older adults activate the agonist and antagonist muscles with greater temporal delay than young adults (Christou et al. 2007). This finding is further supported by studies in young adults that demonstrate that greater coactivation of the antagonistic muscles is used to enhance precision of goal-directed movements (Gribble et al. 2003; Osu et al. 2002) In summary, altered muscle activation may inhibit older adults from extracting task-relevant information during practice and consequently compromise their motor learning capability.

The purpose of this study, therefore, was to determine whether muscle activation differences between young and older adults during practice inhibit the ability of older adults to transfer the task ipsilaterally from the lower to the upper limb. To determine age-associated differences in motor learning that are primarily related with the descending command, we used a rapid goal-directed movement task (Elliott et al. 2010; Yan et al. 2000). We hypothesized that older adults would exhibit impaired ipsilateral transfer compared with young adults due to an altered activation of the antagonistic muscles during practice.

Part of the data has been presented in abstract form (Chen et al. 2012a).

METHODS

Participants

Twenty young (25.1 ± 3.9 yr; 10 women, 10 men) and 20 older (71.5 ± 4.8 yr; 10 women, 10 men) adults volunteered to participate in this study. All subjects reported being healthy without any known neurological impairments and were right-handed (Oldfield 1971) and right-footed (Elias et al. 1998). The Institutional Review Board at University of Florida approved the procedures of this study. All subjects signed a written informed consent form before participating in the study.

Experimental Approach

Figure 1 describes our experimental protocol. Subjects were divided into a practice group and a no-practice control group. The subjects in the practice group participated in two experimental sessions. In the first session, they practiced the goal-directed task with ankle dorsiflexion for 100 trials. In the second session (transfer task), which occurred 24 h later, they performed the goal-directed task with elbow flexion. The subjects in the no-practice group participated in one experimental session. They performed the goal-directed task only with the elbow joint (elbow flexion).

Fig. 1.

Experimental approach. Half of the subjects practiced the goal-directed task with ankle dorsiflexion prior to transferring the task to an ipsilateral elbow flexion movement (practice groups, top). The other half of the subjects only performed the task with elbow flexion and served as a control group (no-practice groups, bottom). Subjects were seated comfortably in an upright position for both tasks. For the ankle task, the left foot was restrained in an ankle device, which allowed only ankle dorsiflexion and plantarflexion movements. For the elbow movements, the left forearm was restrained in an arm device, which allowed only elbow flexion and extension movements. Surface EMGs recorded the muscle activity from agonist (tibialis anterior and biceps) and antagonist (soleus and triceps) muscles for both tasks.

Each session lasted ∼2 h. At the beginning of each session, we explained the experimental procedures and the goal-directed task (ankle or elbow joint movements) to the subjects. Each subject performed the following procedures within a session: 1) maximal voluntary contraction (MVC) tasks with ankle dorsiflexion/plantarflexion and elbow flexion/extension; 2) practice of 3–5 goal-directed movement trials at a target different from the actual target; 3) 100 goal-directed movement trials with either the ankle or the elbow; 4) repetition of the MVC task.

Experimental Arrangement

Experimental setup and apparatus.

Each subject was seated comfortably in an upright position and faced a 32-in. monitor (Sync Master 275t+, Samsung Electronics America) that was located 1.25 m away at eye level. The monitor was used to display the movement produced by elbow flexion and ankle dorsiflexion with a custom-written program in MATLAB (MathWorks, Natick, MA). All subjects affirmed that they could see the display clearly.

For the ankle task the subject's position was as follows: the left hip joint was flexed to ∼90° with 10° abduction, the knee was flexed to ∼45°, and the ankle was plantarflexed to ∼15°. The left foot rested on a customized foot device with an adjustable foot plate and was secured by straps over the metatarsals to ensure a secure position and simultaneous movement between the device and the foot (Fig. 1). The axis of rotation of the customized foot device was positioned in line with the axis of rotation of the left ankle. This arrangement allowed only dorsiflexion and plantarflexion of the ankle. In this study we focused on the tibialis anterior and soleus muscles, which contribute significantly to dorsiflexion and plantarflexion. For the elbow task the subject's position was as follows: the shoulder joint was flexed to ∼90° and horizontal abducted ∼45°, and the elbow was flexed to ∼60°. The left forearm was secured against an adjustable metal plate with a padded strap ∼2–4 in. proximal from the left wrist. The subject held a closed grip around a handle and thus maintained a neutral position of the left wrist (Fig. 1). The axis of rotation of the customized arm device was positioned in line with the axis of rotation of the left elbow. This arrangement allowed only flexion and extension of the elbow. Although other muscles may contribute, in this study we focused on the biceps brachii and triceps brachii muscles. We chose the left leg and arm for this study because they are the nondominant limbs and thus the task would be more novel to both young and older adults compared with the dominant limbs.

Limb position.

The displacement of the ankle (dorsiflexion) was measured with a low-friction potentiometer (SP22G-5K, Mouser Electronics, Mansfield, TX) that was located directly lateral to the fibular malleolus. The displacement of the elbow (elbow flexion) was also measured with a low-friction potentiometer (SP22G-5K, Mouser Electronics) that was located directly lateral to the lateral epicondyle of the humerus. The ankle and elbow position signals were sampled at 1,000 Hz with a NI-DAQ card (model USB6210, National Instruments, Austin, TX) and stored on a personal computer.

Force.

The maximum voluntary force exerted during ankle dorsiflexion and plantarflexion as well as elbow flexion and extension was measured with a force transducer (model 41BN, Honeywell, Morristown, NJ) that was located in parallel with the force direction on the customized foot and hand devices. The ankle and elbow force signals were high-pass filtered at 0.03 Hz, amplified 50 times (Bridge-8, World Precision Instruments), sampled at 1,000 Hz with a NI-DAQ card (model USB6210, National Instruments), and stored on a personal computer.

EMG measurement.

Muscle activation was recorded with a Trigno wireless EMG system (Delsys, Boston, MA) from the biceps brachii and triceps brachii muscles during the elbow flexion task and from the tibialis anterior and soleus muscles during the ankle dorsiflexion task. The recording electrodes on each muscle were placed on the skin and in line with the muscle fibers. The location for each electrode was selected according to the European Recommendations for Surface Electromyography (Hermens et al. 2000). The EMG signals were band-pass filtered from 20 to 450 Hz, amplified 1,000 times, sampled at 1,000 Hz with a NI-DAQ card (model USB6210, National Instruments), and stored on a personal computer.

MVC task.

For the ankle joint, we identified the MVC for ankle dorsiflexion and ankle plantarflexion. Subjects increased force to their maximum in 3 s and maintained the maximal force for ∼3 s. They exerted three to five MVCs or until two MVC trials were within 5% of each other. One-minute rests were provided between consecutive trials. For the elbow joint, we identified the MVC for elbow flexion and elbow extension. MVC force for the elbow joint was quantified in the same fashion as that for the ankle joint. MVC tasks were repeated at the end of the session to assess whether the experimental task induced muscle fatigue. In addition, we recorded the peak EMG (average of 0.5 s around the peak EMG of the trial) during MVC, which was used to normalize the EMG during the goal-directed movements.

Goal-directed movements.

Subjects performed goal-directed movements that involved accurately matching the peak displacement of the limb to a target by performing ankle dorsiflexion or elbow flexion. The movements were unloaded, and the target displacement was 9° for the ankle and 18° for the elbow. Therefore, the target displacement for each limb was ∼13% of the available range of motion for each joint. The time given to reach the target was 180 ms for both movements.

The task contained the following three phases (Fig. 2A): 1) GET READY, 2) MOVE, and 3) FEEDBACK. The GET READY phase began by the presence of a red target on the monitor for 2 s. This was a cue for the subjects to be ready for the MOVE phase. The MOVE phase began when the red target switched to a green target. This target color change was the cue for subjects to perform the goal-directed movement. The green target stayed on the monitor for 3 s, and subjects were instructed to perform a movement at their convenience (not a reaction time task). The recording of the task began when the subject initiated the movement within the 3 s of the MOVE phase. The FEEDBACK phase began at the end of each MOVE phase and lasted for 5 s. We provided the subjects with visual feedback of their movement trajectory trace relative to the targeted angle-time end point (Fig. 2B). The visual gain was kept constant at 1° (visual angle) for all trials (Christou 2011; Vaillancourt et al. 2006).

Fig. 2.

Goal-directed movements. A: the task was divided into the following 3 phases: GET READY, MOVE, and FEEDBACK. During the GET READY phase, subjects viewed a red target on the monitor for 2 s and remained relaxed. The subjects initiated the movement (MOVE; no reaction was required) when the target switched color from red to green. This phase lasted for 3 s. At the end of the MOVE phase, visual feedback was provided for 5 s to the subjects about their movement trace relative to the target (FEEDBACK phase). B: we quantified the end point error of each movement as the shortest distance of the movement end point to the target. The end point error was the resultant error of the normalized position and timing errors (see methods for more details).

Data Analysis

Data were analyzed off-line with custom-written programs in MATLAB (MathWorks, Natick, MA). We calculated the end point accuracy, end point variability, and activation of the agonist and antagonist muscle during the ankle dorsiflexion and elbow flexion goal-directed tasks. We eliminated a very small number of trials (<5% of the total trials) on the basis of the following two criteria: 1) the end point error was beyond ±3 standard deviations of the mean performance; 2) the peak position was <10% of the targeted position.

Movement Control

MVC force.

The MVC force was defined as the highest force exerted with ankle dorsiflexion/plantarflexion and elbow flexion/extension during the MVC task.

End point error.

To calculate the end point error we quantified the position and time errors. Position error was quantified as the absolute deviation from the targeted peak displacement, whereas time error was quantified as the absolute deviation from the targeted time to peak displacement. The hypotenuse between the position error and time error distance represents the shortest distance from the target and thus the end point error (Fig. 2B). To be able to quantify the hypotenuse, we normalized the position and time errors to have the same units (%). The position error was normalized to the targeted peak displacement (Eq. 1) and the time error was normalized to the targeted time to peak displacements (Eq. 2). The end point error was quantified as the hypotenuse of the position and time errors (Eq. 3).

$$\text{Position error}\hspace{0.17em}\left(\text{\%}\right)=\frac{\text{peak displacement}\hspace{0.17em}(°)}{\text{targeted peak displacement}\hspace{0.17em}(°)}\times 100$$ (1)

$$\text{Time error}\hspace{0.17em}\left(\text{\%}\right)=\frac{\text{time to peak displacement}\hspace{0.17em}\left(\text{ms}\right)}{\text{targeted time to peak displacement}\hspace{0.17em}\left(\text{ms}\right)}\times 100$$ (2)

$$\text{End point error}\hspace{0.17em}\left(\text{\%}\right)=\sqrt{{\left(\text{time error}\right)}^2+{\left(\text{position error}\right)}^2}$$ (3)

End point variability.

We quantified position and time variability with the coefficient of variation (CV; standard deviation/mean × 100) of the peak displacement (position) and time to peak displacement (time). To quantify the CV for position and time we calculated the mean and standard deviation of the peak displacement and time to peak displacement for each block of practice.

Neuromuscular Activation

Agonist and antagonist EMG activity.

The interference EMGs were rectified and smoothed with a fourth-order Butterworth digital filter with a cutoff frequency of 6 Hz (Christou et al. 2007; Poston et al. 2008a, 2008b, 2010) (Fig. 3). This filter was used to identify the amplitudes, onsets, and offsets of the EMG bursts for the primary agonist and antagonist muscles of the elbow and ankle, respectively.

Fig. 3.

Muscle activation. We quantified the EMG bursts from the agonist and antagonist muscles by detrending, rectifying, and 6-Hz low-pass filtering the interference EMGs. Next, we determined the start, end, and peak of each burst and extracted the following: 1) EMG burst duration, 2) EMG burst amplitude, 3) EMG burst, 4) EMG temporal relation, 5) EMG coactivation, 6) coefficient of variation (CV) of EMG burst duration, 7) CV of EMG burst amplitude, 8) CV of EMG burst, 9) CV of EMG temporal relation, and 10) CV of EMG coactivation. See methods for detailed explanation of each of these variables.

Muscle activity.

We examined muscle activity by quantifying the EMG bursts in the following ways.

The EMG burst duration was the time between EMG onset (>15% of the peak EMG) and EMG offset (>15% of the peak EMG).

The EMG burst amplitude was the peak EMG activity within the EMG burst duration normalized to the EMG recorded during the MVC.

The EMG burst was defined as the sum of the EMG values within the EMG burst duration (area under the EMG burst curve) normalized to the sum of EMG recorded during 360 ms around the peak EMG of the MVC task.

The agonist-antagonist EMG temporal relation was defined as the time difference between the time of the peak EMG burst amplitude for the agonist and antagonist muscles.

The agonist-antagonist EMG coactivation was defined as indicated in Eq. 4 (Falconer and Winter 1985).

$$\text{EMG coactivation}\hspace{0.17em}\left(\text{\%}\right)=\frac{2\times \text{antagonist EMG burst amplitude}}{\text{agonist}+\text{antagonist EMG amplitude}}\times 100$$ (4)

Variability of muscle activity.

We determined the variability of muscle activity by examining the CV of the EMG burst characteristics mentioned above for each block of practice.

Statistical Analysis

The major dependent variables were 1) position error, 2) time error, 3) end point error, 4) CV of peak displacement, 5) CV of time, 6) EMG burst duration, 7) EMG burst amplitude, 8) EMG burst, 9) EMG temporal relation, 10) EMG coactivation, 11) CV of EMG burst duration, 12) CV of EMG burst amplitude, 13) CV of EMG burst, 14) CV of EMG temporal relation, and 15) CV of EMG coactivation. A two-way mixed ANOVA [2 age (young vs. older) × 5 blocks (20 trials/block)] with repeated measures on blocks was used to compare all parameters during task acquisition (ankle dorsiflexion). A three-way mixed ANOVA [2 age (young vs. older) × 2 groups (practice and no-practice control groups) × 5 blocks (20 trials/block)] with repeated measures on blocks was used to compare all parameters for the ipsilateral transfer task (elbow flexion).

We used stepwise linear regression analysis to establish statistical models that could predict movement performance from muscle activity. The goodness of fit of each model was given by the squared multiple correlations (R²; Green and Salkind 2004). All statistical analyses were performed with the IBM statistics 21.0 statistical package (IBM, New York, NY). The α level for all statistical tests was 0.05. Data are reported as means ± SD in the text and as means ± SE in figures. Only the significant main effects and interactions are presented, unless otherwise noted.

RESULTS

MVC Force

The MVC force for ankle dorsiflexion and elbow flexion was not significantly different between young and older adults (ankle: 104.52 ± 24.73 vs. 82.22 ± 27.12 N, P = 0.07; elbow: 132.07 ± 59.18 vs. 107.54 ± 62.99 N, P = 0.21). Furthermore, the two age groups exhibited similar MVC forces before and after the experimental session (ankle: 93.37 ± 27.73 vs. 97.34 ± 26.16 N, P = 0.99; elbow: 119.81 ± 61.59 vs. 119.89 ± 59.49 N, P = 0.94).

Task Acquisition

Older adults exhibited greater ankle dorsiflexion end point error and variability during task acquisition. This is demonstrated in Fig. 4A, which shows the end points from 100 ankle dorsiflexion trials for one young and one older adult. Practice of the ankle task decreased position error (F_4,72 = 15.46, P < 0.001), time error (F_4,72 = 4.149, P = 0.004), and end point error (F_4,72 = 10.33, P < 0.001; Fig. 4B). Similarly, the position variability (F_4,72 =13.25, P < 0.001) and time variability (F_4,72 = 3.21, P = 0.018) also decreased after practice. The rate of improvement in end point error and variability with practice was similar for the two age groups (P > 0.1 for all age × block interactions). In addition, we found that the soleus EMG amplitude (F_4,72 = 5.99, P < 0.001), soleus EMG burst (F_4,72 = 3.36, P = 0.014), and EMG coactivation (F_4,72 = 7.1, P < 0.001) decreased with practice. The rate of decrease in all other EMG parameters was similar for young and older adults (P > 0.1 for all age × block interactions).

Fig. 4.

Task acquisition with ankle dorsiflexion. A: representative ankle end point data for a young and an older adult. Filled circle and triangle symbols represent the mean and SD of the movement end point of 100 trials for the young and older adults, respectively. The end point error and end point variability for both the position and time components of the task were greater for the older adult. B: overall data for ankle dorsiflexion end point error for young and older adults. Gray and black lines represent the raw (smoothed) performance of young and older adults, respectively. Open and filled circles represent the average performance of 20 trials for young and older adults, respectively. Both age groups decreased end point error with practice, and their rate of improvement was similar. The end point error was greater for older adults.

Movement control.

On average, older adults exhibited greater time error (F_1,18 = 6.86, P = 0.017) and end point error (F_1,18 = 4.84, P = 0.04; Fig. 5A) compared with young adults. In contrast, the position error was not significantly different between the two age groups (F_1,18 = 0.737, P = 0.4). Older adults exhibited greater time variability compared with young adults (F_1,18 = 6.93, P = 0.017; Fig. 5B). The position variability, however, was not significantly different between the two age groups (F_1,18 = 1.82, P = 0.194).

Fig. 5.

Movement performance and muscle activity during task acquisition. Older adults exhibited greater end point error (A) and greater time variability (B) than young adults. Similarly, older adults exhibited greater tibialis anterior (TA) muscle EMG burst (C) and greater soleus muscle EMG burst (D) than young adults. *Significant differences (P < 0.05) between young and older adults.

Muscle activity.

The neural activation of the tibialis anterior and soleus muscles was different for young and older adults during task acquisition. Specifically, we found that older adults exhibited greater tibialis anterior EMG burst (F_1,18 = 4.92, P = 0.04; Fig. 5C) and soleus EMG burst (F_1,18 = 4.43, P = 0.049; Fig. 5D). The EMG amplitude of the tibialis anterior (F_1,18 = 7.09, P = 0.016) and soleus (F_1,18 = 5.09, P = 0.037) muscles was significantly different between young and older adults. Furthermore, the soleus EMG amplitude variability was greater in older adults compared with young adults (F_1,18 = 7.18, P = 0.015). The agonist-antagonist EMG temporal relation was significantly different for young and older adults (F_1,18 = 15.3, P < 0.001), indicating that older adults exhibited a longer time between the antagonistic muscle activities. All other EMG parameters were not significantly different between young and older adults (P > 0.05).

Ipsilateral Transfer

Movement control.

There was a significant age × group interaction for elbow end point error (F_1,36 = 5.20, P = 0.029; Fig. 6A). Young adults who practiced with ankle dorsiflexion prior to performing the task with the elbow flexors exhibited significantly lower end point error during the transfer task (t = 3.819, P = 0.001) than young adults who did not practice with their ankle. In contrast, older adults who practiced ankle dorsiflexion prior to performing the task with the elbow flexors exhibited end point error similar to the older adults who received no ankle practice (t = 0.967, P = 0.346). The young adults who received ankle practice exhibited 30% lower end point error for the elbow flexion task compared with older adults who also received similar ankle practice (t = 3.721, P = 0.002).

Fig. 6.

Movement performance and muscle activity during the transfer task. A: the elbow end point error was similar for older adults who received ankle practice and those who did not. In contrast, young adults who received ankle practice exhibited lower elbow end point error compared with young adults who did not receive any training. Most importantly, young adults who practiced the task with the ankle exhibited ∼30% lower end point error with the elbow compared with older adults who received equal ankle practice. B: interestingly, the results were similar for the antagonist muscle activity during the transfer task. The triceps EMG burst was similar for older adults who received ankle practice and those who did not. In contrast, young adults who received ankle practice exhibited lower triceps EMG burst compared with the young adults who did not practice. Finally, young adults who practiced the task with the ankle exhibited ∼60% lower triceps EMG burst activity compared with older adults who received equal ankle practice. *Significant differences (P < 0.05) between young and older adults in experimental group; ^#significant differences (P < 0.05) between control and experimental groups in young adults.

Muscle activity.

There was a significant age × group interaction for triceps EMG burst (F_1,36 = 4.62, P = 0.038; Fig. 6B). We found that young adults who practiced ankle dorsiflexion prior to performing the task with the elbow flexors exhibited significantly lower triceps EMG burst during the transfer task (t = 2.82, P = 0.011) compared with young adults who did not practice with their ankle. In contrast, older adults exhibited similar triceps EMG burst with or without prior ankle practice (t = 0.54, P = 0.596). The young adults who received ankle practice exhibited 60% lower triceps brachii activity for the elbow flexion task compared with older adults who also received similar ankle practice (t = 2.838, P = 0.011).

The above results suggest that only young adults were able to transfer movement accuracy from the ankle to the elbow, likely because of their ability to transfer the antagonist muscle activity. Therefore, we examined the association between the soleus muscle (antagonist during ankle dorsiflexion movements) during task acquisition and the triceps brachii (antagonist during elbow flexion movements) during the transfer task. The soleus EMG burst during acquisition (ankle dorsiflexion) was positively related (R² = 0.31, P = 0.01; Fig. 7) with the triceps EMG burst during the transfer task (elbow flexion). Subjects who exhibited lower soleus EMG burst during task acquisition (mostly young adults) exhibited lower triceps EMG burst during the transfer task. Furthermore, we examined the relation between the practice-induced decrease in soleus EMG burst and decrease in end point error (Fig. 8) to determine whether decreased antagonist muscle activity induced better task performance during task acquisition in young and older adults. Interestingly, we found that the practice-induced decrease in soleus EMG burst was positively related to the decrease in ankle movement error for young adults (R² = 0.65, P = 0.005) but not in older adults (R² = 0.10, P = 0.37).

Fig. 7.

Association between the antagonist muscle activity during task acquisition and transfer. There was a significant association (R² = 0.31, P = 0.023) between the soleus EMG burst (acquisition task) and the triceps EMG burst activity (transfer task). Subjects who exhibited a lower soleus EMG burst during task acquisition (primarily young) also exhibited a lower triceps EMG burst during the transfer task.

Fig. 8.

Association between the antagonist muscle activity and movement accuracy during task acquisition. There was a significant association between the practice-induced decrease in ankle end point error and the decrease of the soleus EMG burst in young adults (R² = 0.65, P = 0.005) but not in older adults (R² = 0.1, P = 0.37). Young subjects who exhibited the greatest decrease in ankle end point error with practice had the greatest decreases in soleus EMG burst activity.

DISCUSSION

The focus of this study was to determine whether differential activation of the agonist and antagonist muscles during practice compromises motor learning in older adults relative to young adults. We studied motor learning by examining the ipsilateral transfer of a rapid goal-directed task from ankle dorsiflexion to elbow flexion. Only young adults who received practice of the task with ankle dorsiflexion exhibited greater movement accuracy with elbow flexion movements. The enhanced ability of young adults relative to older adults to transfer the task to the ipsilateral upper limb appears to be related to their capacity to transfer the pattern of muscle activation. Specifically, the antagonist muscle burst seems to be of importance for the goal-directed task we used in this study. These results provide novel evidence that the deficient motor learning in older adults, demonstrated by impaired ipsilateral transfer, may be related to a differential activation of the antagonist muscle during practice.

Ipsilateral Transfer

Motor learning is often examined with a transfer task rather than a retention task, especially to compare two groups with different performance capabilities (Kantak and Winstein 2012; Seidler 2010). Performance differences during the retention task may be due to differences in the general ability of the two groups to perform the task and independent of the amount that has been learned during practice. A possible alternative is to examine the relative improvements from the initial stages of practice to the retention task. In that case, the greater improvements will occur in the group with the lowest performance capabilities. However, the greater improvements may be due to the fact that the group with lower capabilities has more room to improve after practice. The only time a retention task will be useful in providing meaningful information about motor learning will be when the performance of the two groups is similar at the initial stages of practice. Thus, for our study, it was more reasonable to use a transfer task because the initial performance capability was greater for young adults than older adults. In addition, transfer provides information about whether the learned task can be generalized to other effectors, which indicates the flexibility of motor memory (Kantak and Winstein 2012).

Most aging studies on motor learning have incorporated a contralateral transfer paradigm (Hinder et al. 2011; Panzer et al. 2011; Seidler 2006; Wang et al. 2011). Because of the likelihood of interhemispheric interaction during practice (Hinder et al. 2011; Wang et al. 2011), we used an ipsilateral transfer paradigm to eliminate the potential neural transfer during practice. Despite the paradigm differences, our results support the findings from the contralateral transfer studies and demonstrate that older adults exhibit an impaired ability to transfer motor tasks compared with young adults. This is demonstrated in Fig. 6A. Only young adults who received ankle practice prior to elbow flexion exhibited ∼27% better movement accuracy than young adults who did not receive any practice. In contrast, the movement accuracy of elbow flexion was similar for older adults who received prior ankle practice and older adults who did not receive any practice.

In contrast to the ipsilateral transfer findings, young and older adults exhibited a similar rate of adaptation during ankle dorsiflexion practice (Fig. 4B). This similarity in adaptation is supported by several previous findings (Bo et al. 2009; Chen et al. 2012b; Panzer et al. 2011). For example, Seidler (2006) reported that older adults exhibit a similar adaptation curve and retention performance compared with young adults during a sequential movement task. Furthermore, Panzer et al. (2011) found that although older adults cannot transfer a sequential movement from one arm to another, their ability to adapt to the task was equivalent to young adults. These findings, therefore, indicate that older adults are able to adapt acutely during task practice but appear unable to perform the task in different contexts. The compromised transfer in older adults compared with young adults is likely related to their inability to extract similar information during practice.

There are three potential explanations of why older adults exhibit a similar rate of adaptation but an impaired transfer. First, the task-relevant information during practice may be greater for older adults because they exhibit greater movement errors (Torres-Oviedo and Bastian 2012). Furthermore, error correction may activate brain circuits in older adults (e.g., cortico-striatal) that are healthy but are not necessarily involved in the learning of a motor task (Seidler et al. 2010). In contrast, greater movement error may decrease their ability to form strong neural connections that represent the practiced task and consequently impair their ability to memorize and transfer the task (Wolpert et al. 2011). The greater movement errors during practice for the older adults, therefore, could have interfered with their ability to extract task-relevant information used for generalization. Indeed, this idea is supported by the theoretical framework of Wolpert and colleagues (Wolpert et al. 2011), which proposes that reduced movement performance during practice leads to impaired motor learning. Second, the brain activation is different for young and older adults during task acquisition (Seidler 2010). A common finding is that older adults activate more brain areas to perform the task than young adults (Heuninckx et al. 2005; Mattay et al. 2002). Therefore, short-term adaptation may be enhanced by the recruitment of additional brain areas (Heuninckx et al. 2008; Wu and Hallett 2005). In contrast, long-term retention of the task and the ability to generalize (transfer) may be impaired by the increased activation of the brain. Possibly, the connections among different brain areas may be impaired with increased areas being active (Taniwaki et al. 2007) and consequently limit the ability of older adults to form a strong representation of the task (Morcom et al. 2007). Finally, another possibility may be the deficient reciprocal inhibition in older adults compared with young adults (Hortobagyi et al. 2006; Kido et al. 2004). The impaired activation of the antagonist muscle in older adults during practice, which contributes to their impairments in transferring the task, may be related to age-related loss of cortical reciprocal inhibition (Hortobagyi et al. 2006) and spinal reciprocal inhibition (Kido et al. 2004).

Ipsilateral Transfer and Muscle Activity

It is well known that brain activation differs for young and older adults while practicing a new task and consequently interferes with their ability to transfer a motor task (for a review see Seidler 2010). Nonetheless, it is unknown whether altered muscle activation during practice in older adults also contributes to their impaired ability to transfer a motor task. Our results provide novel evidence that the antagonist muscle activation is critical during practice (Fig. 4B and Fig. 5) and for ipsilateral transfer from ankle dorsiflexion to elbow flexion (Fig. 6B). Specifically, the activity of the antagonist muscle during elbow flexion movement was lower by ∼55% in young adults who received previous ankle practice compared with young adults who did not receive any practice. In contrast, the activity of the antagonist muscle during elbow flexion was similar for older adults who received ankle practice and older adults who did not receive any practice. The differential activation of the antagonist muscle findings are parallel to those observed for movement error.

Furthermore, the importance of lower antagonist muscle activity during the rapid movement task is demonstrated during ankle dorsiflexion practice. With practice, both age groups reduced movement error and the activity of the antagonist muscle (soleus) during ankle dorsiflexion movements. An important finding, however, was that the practice-induced reductions in ankle movement error were significantly related to the reduction in antagonist muscle activity in young (R² = 0.65; Fig. 8) but not in older (R² = 0.10; Fig. 8) adults. This finding suggests that only young adults were able to link the reduction in antagonist muscle activity to the performance of the practiced task. Thus the age-associated difference in ankle movement accuracy observed during task acquisition (Fig. 5) may be related to the activity of the antagonistic muscles.

Our previous findings (Christou et al. 2007) support the notion that impaired end point performance in older adults is related to altered activation of the agonist and antagonist muscles. It appears that during fast goal-directed contractions young adults activate the agonist and antagonist muscles closer together than older adults. Our present results, which are based on fast goal-directed movements, also demonstrate that the temporal relation of the agonist and antagonist muscles is closer for young adults than for older adults. This suggests that young adults use this closer temporal antagonistic muscle activation as a strategy to minimize end point error during goal-directed movements. Indeed, work by Gribble et al. (2003) suggests that when the target becomes more challenging for young adults (smaller in size) they adapt a similar strategy to their antagonistic muscles.

Higher Centers and Impaired Ipsilateral Transfer in Older Adults

Our outcomes, which focus on the motor output and muscle activation, strongly suggest that the observed age-associated differences in transfer and muscle activity are due to a differential activation of higher centers. This finding is supported by the rapid goal-directed movement task selected for this study. Rapid movements are largely unaffected by online sensory and visual feedback (Elliott et al. 2010; Yan et al. 2000). Therefore, the age-associated differences in motor learning that are demonstrated here are likely due to differences in the cortical input to the motor neuron pool (motor program/motor planning) rather than an online deficiency in sensory or visual feedback. Indeed, there is evidence that transfer is correlated with activity in the right cingulate gyrus, left superior and right inferior parietal lobule, left middle occipital gyrus, and bilateral cerebellum (Seidler 2010). An alternative or even an additional mechanism that contributes to the differences in the ipsilateral transfer between young and older adults may be the age-associated loss in reciprocal inhibition. This deficit may influence the activation of the antagonist muscle during practice of the task and consequently inhibit task acquisition (see explanation in paragraphs above).

Considerations

The sensitivity of the surface EMG could partially limit us from examining the contribution of all the neuromuscular mechanisms that could have contributed to the impaired motor learning in older adults. Although our results indicate that the antagonist muscle activation predicts the impaired transfer in older adults, other muscles or components of the EMG may also have contributed. Specifically, the brachioradialis muscle, which was not examined in this study, contributes significantly to elbow flexion in our movement task (Hunter et al. 2003). Future studies should examine in more detail how all the agonist and antagonist muscles are activated differently in young and older adults during practice and transfer tasks. Most importantly, we should examine the activation of multiple motor units and how it affects the ability to transfer. The motor unit is the final common pathway from the nervous system to the muscle (Duchateau and Enoka 2011; Sherrington 1925) and may provide significant insight on how the nervous system plans muscle activity differently for young and older adults after practice. Finally, our results are limited to rapid goal-directed movements. It may be interesting to examine slower tracking movements. Such movements are influenced by online sensory feedback, which is often impaired in older adults (Kennedy and Christou 2011), and thus may exacerbate the age-associated differences in transfer of motor tasks.

In summary, we provide evidence that motor learning is compromised in older adults, as demonstrated by impaired ipsilateral transfer. This impairment was related to the inability of older adults to activate the antagonist muscle similarly to young adults during practice. These results provide original evidence that the compromised motor learning in older adults is related to their altered muscle activation during practice, likely interfering with the extraction of task-relevant information.

GRANTS

This work was supported by National Institute on Aging Grant R01 AG-031769 to E. A. Christou.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: Y.-T.C., M.K., and E.A.C. conception and design of research; Y.-T.C., M.K., and E.A.C. performed experiments; Y.-T.C., M.K., and E.A.C. analyzed data; Y.-T.C., M.K., E.J.F., and E.A.C. interpreted results of experiments; Y.-T.C., M.K., and E.A.C. prepared figures; Y.-T.C., M.K., E.J.F., and E.A.C. drafted manuscript; Y.-T.C., M.K., E.J.F., and E.A.C. edited and revised manuscript; E.A.C. approved final version of manuscript.