Effect of simple motor exercise on motor adaptation in complex dynamic tasks; Exploring age-related variations

Abstract

Dexterous manual actions rely on the integration of precise sensorimotor control and adaptive learning. However, it remains unclear how repetition of simple motor tasks influences subsequent adaptation to force perturbations in a dynamic manipulation task. This study examined whether different types of motor repetition, force-based or movement-based, affect the adaptation process in young and older adults. Sixty right-handed participants (30 young, 30 older) performed a dynamic manipulation task using a robotic interface, where they had to counteract perturbation torques during a handle-lifting movement. Before the perturbation trials, participants engaged in either force repetition, which required producing discrete isometric torque pulses, or movement repetition, which involved continuous wrist rotations. We found that young adults who performed force repetition exhibited enhanced adaptation in the dominant hand, while movement repetition did not yield the same benefit. However, older adults showed no significant modulation of adaptation based on repetition type. Additionally, across all participants, adaptation performance differed between supination and pronation directions, with greater accuracy observed in the supination condition. This asymmetry was more pronounced in young adults and in the non-dominant hand. These findings suggest that the benefits of force repetition for predictive motor control may depend on both age and limb dominance, with implications for motor training and rehabilitation strategies.

Graphical Abstract

NEW & NOTEWORTHY

This research presents a novel approach to studying how motor repetition influences adaptation in dexterous manipulation. Departing from traditional reaching tasks, we used a dynamic manipulation paradigm where participants were adapting to dynamic perturbations after force- or movement-based repetition exercises. These tasks engaged predictive or feedback motor processes. Our design accounted for hand dominance and age-related differences, providing a comprehensive framework to investigate how prior motor experience shapes learning, adaptation, and transfer in hand control.

INTRODUCTION

Dexterous manipulation is an essential component of the activities of daily living, which can be achieved through skilled coordination and precise sensorimotor control of the hands. Any changes in the environmental dynamics can disrupt the central nervous system’s (CNS) proficiency in executing motor actions that were previously practiced extensively (1). However, the CNS can adapt by refining the feedforward predictive motor commands in response to the changes or perturbations and improve the execution of well-practiced movements through repeated exposure to specific tasks (2). Extensive research has been conducted on learning and motor adaptation in the upper extremities, with a primary focus on the simple reaching movements in a novel environment, such as a curl field or visuomotor rotation (3). The ability to gradually produce more appropriate predictive motor commands in these tasks is supported by several parallel mechanisms (4). Among these mechanisms, the most extensively studied one is the model-based error-driven adaptation, which attempts to update the motor commands via error signals from previous exposures to the same context (5, 6). More recent research has shown that error signals can drive motor adaptation both implicitly and explicitly: the implicit component recalibrates the internal models of the body and/or environment without conscious awareness, whereas the explicit component involves cognitive strategies (7–9). In contrast, there are alternative mechanisms that do not require error signals but rather depend on repeating the same action, i.e., use-dependent learning (10), or receiving rewards, i.e., reinforcement learning (11). Although studies using dexterous manipulation tasks have been rare, it has been found that adaptation to novel object dynamics can also be attributed to both error-dependent and error-independent processes (Fu and Santello 2015).

An important question in motor learning is how adaptation to a specific change can generalize to a different one (2), given its relevance in clinical applications such as rehabilitation (13). Numerous studies with reaching movements have demonstrated that the extent to which motor adaptation in one context transfers to another depends on many factors, such as target similarity (14), limb postures (15, 16), end-effectors (17, 18), and perturbation types (19). Generalization in dexterous manipulation tasks also showed similar results (20–22). However, it remains unclear how different adaptation mechanisms contribute to the various degrees of generalization across different tasks, as most studies have focused on tasks that rely on error-dependent processes that share similar error signals. Interestingly, there is scattered evidence showing that cross-task generalization is possible even when two tasks do not share similar error signals. For instance, forcefield compensation in free motion has been shown to transfer from isometric force training without proprioceptive errors (23). Similarly, research on grip force scaling has demonstrated that the grip component of an object-lifting task following finger-force pinching was higher than lifting the same object without a preceding pinching (24). These findings suggested that repeated exposure to specific motor actions can prime the sensorimotor system for related tasks, highlighting the potential generalization from error-independent learning, e.g., use-dependent learning, to a subsequent motor learning process that is typically dominated by error-based motor adaptation.

Another important question is how the multi-process motor adaptation and the generalization of motor adaptation may change with age. It is well-established that aging significantly influences motor adaptation (25). Notably, older adults (OA) consistently show slower adaptation rates and lower asymptotic performance under novel visuomotor perturbations, yet they exhibit comparable or even larger aftereffects once the perturbation is removed (Bock2005b; Fernández-Ruiz et al. 2000; Hegele and Heuer 2010). It was theorized that this dissociation suggested that aging only impairs the explicit but not the implicit component of the error-based adaptation (25). During exposure to perturbations, young adults (YA) benefited from explicit strategies more than OA (29), but the explicit strategies play a less significant role during aftereffects (Taylor et al., 2014). Indeed, a recent study directly demonstrated that recalibration of internal models and short-term retention of motor memories are intact in OA during reaching tasks with visuomotor rotation (30). In contrast, they observed a decline of the explicit component of adaptation, which may be associated with a reduced capacity of visuospatial working memory (31). Nevertheless, existing studies have focused on error-dependent mechanisms (both explicit and implicit), but no studies have examined how age may alter the error-independent motor adaptation mechanisms and their interaction with error-dependent motor adaptation.

The present study examined how use-dependent learning may affect overall motor learning of a dynamic task in both young and older adults. We chose use-dependent learning, a form of error-independent motor adaptation for predictive control, mainly because it can be elicited through motor repetitions of simple hand exercises that can be easy to implement in clinical settings. We also designed the dynamic task to resemble real-world tasks that require simultaneous grip force control and wrist torque generation, e.g., lifting, rotating, or stabilizing handheld objects. Specifically, using a robotic interface, participants performed a dexterous manipulation task to adapt to imposed perturbation torques after undergoing one of the two types of repetition exercise trials. Both exercise types promoted the use-dependent adaptation mechanism and did not provide error feedback related to the main manipulation task. However, these two exercise types differed in what motor output is repeated: one type focused on isometric hand force production, whereas the other emphasized hand motion. Because motor adaptation in the dynamic task for participants to adapt required the production of hand force to overcome the perturbation, we anticipated that the use-dependent learning of hand force would be more beneficial than the use-dependent learning of hand motion. Therefore, we hypothesized that in YA, force repetition would enhance motor adaptation more effectively than movement repetition. Additionally, we assessed motor adaptation as a predictive feedforward control process for both right and left hands in this study. Because the dynamic dominance framework states that two limbs adapt to perturbation differently and the dominant arm relies more on feedforward control (32), we further expected the effect of use-dependent learning to be more pronounced in the dominant hand, reflecting such lateralized motor learning advantages. Lastly, OA may have declined capacity in predictive control during dynamic object manipulation tasks (33), reducing the benefit of use-dependent learning. Therefore, we hypothesized that the differential effect of repetition exercise types may be less pronounced in OA.

MATERIALS AND METHODS

Participants

A total of 60 Healthy right-handed individuals, without any history of neuromuscular disease or physical injury to the upper body within the past two years, participated in this research. There were 30 young adults (YA, 16 female, age 26.9 ± 6.9 years), and 30 older adults (OA, 26 female, age 62.6 ± 5.2 years). The sample size was determined with a priori power analysis with alpha 0.05, beta 0.8, and a moderate to large effect size. The young adults were recruited locally around the campus of University of Central Florida, whereas the older adults were recruited from local communities of Orlando. A laterality index (LI) was measured for each participant using an adapted version of the Edinburgh Handedness Inventory (34). The LI scores for YA and OA groups were 62.6 ± 45.7 and 80.9 ±18.8, respectively. Note that the LI variability was greater in the younger group, but we did not find a significant difference between age groups (Welch’s t-test). The larger variability of LI scores in older adults is consistent with previous studies (35), which may be explained as a generational effect (36). All participants had normal or correct-to-normal vision. They were naïve to the purpose of the study and gave their informed consent. The experimental protocols were approved by the Institutional Review Board at the University of Central Florida in accordance with the Declaration of Helsinki.

Dynamic manipulation task

The dynamic manipulation task used two haptic robotic devices (Phantom 1.5HF, 3D Systems) located on the table in front of a vertically placed computer monitor. The distance between the screen and participants’ eyes was approximately 1 m. The end-effectors of the robots were attached to the sides of a custom 3D-printed frame that hosts a rectangular handle (Fig.1A). Participants were required to grasp the handle placing their thumb at the middle of one side and the other four fingers on the opposite side. The movement of the handle was recorded through the robots at a sampling rate of 500 Hz. The handle’s movement was constrained to a vertical plane (x-y) that was aligned with the participants’ frontal plane, enabling them to move their hands comfortably in the vertical direction. The constraints were implemented with a soft virtual spring-damper (spring constant, K_w = 0.25 N/mm; damping constant, B_w = 0.03 N·s/mm). The dynamic manipulation task was implemented with the CHAI3d software library to control the robots and render visual feedback (37). Although participants could see their hands and the robots in their peripheral vision, they were instructed to concentrate on the visual feedback shown on the screen to guarantee accurate task execution.

Figure 1.

A. Experimental Setup. The red dots denote the connection points to the robots. B. Visual display of the manipulation task. Handle rotation Ө was mapped to the green cursor movement in the x direction. The dashed line represents a sample trajectory of the cursor, which was not visible to the participants during the experiment. Blue rectangle and grey circle are the ‘Target’ and ‘Start’ zones, respectively. C. Generation of viscous force dynamics. V represents the velocity of the reference point, and F is proportional to V. Note that the left and right hands must produce compensatory torques in the opposite direction for the same direction of dynamics (CW or CCW).

In every trial, participants were instructed to move a cursor from a “Start” zone (gray circle) into a target zone (blue rectangle; Fig. 1B), while minimizing the horizontal movement of the cursor. The cursor was controlled by the movement of the handle. Specifically, the vertical position of the cursor corresponded to the vertical movement of the handle, whereas the horizontal position of the cursor was controlled by the handle rotation Ө (Fig. 1C). A horizontal gray line moves vertically alongside the cursor. The horizontal displacement of the cursor was indicated as a solid green bar on the gray line, representing the range of handle rotation throughout the entire trial. A trial was considered successful if the cursor ended in the target zone within a time window of 450–550 ms after leaving the Start zone, while the green bar remained smaller than the width of the target zone. This represents a ±5° error margin for handle rotation. The bottom and top of the target zone corresponded to 160 mm and 180 mm vertical movement of the handle from the center of the start zone. If participants meet both speed and precision requirements, the target turns green as “Success”. Otherwise, the target turns red as “Failure”. Participants are instructed to get as many successful trials as possible regardless of changes in the dynamics of the task.

There were three types of dynamics, which were implemented by the robots producing viscous forces on the handle (Fig. 1C). The force F was proportional to the velocity V of specific reference points: F = BV. In Null conditions, the reference point is at the center of the handle, and the resistive force was evenly distributed between two robots. Such forces do not generate torques that perturb the orientation of the handle, therefore requiring zero compensatory torque from the participants. In contrast, the right and left reference points not only provided a resistive lifting force but also introduced a perturbation torque in the clockwise (CW) or counterclockwise (CCW) directions, respectively, by producing F with a single robot. Consequently, subjects had to adapt to these perturbations and counteract with compensatory torques T_C during handle lifting to successfully complete the task. For example, when a participant lifts the handle with the right hand while a CW perturbation is applied, the perturbation would cause the handle to rotate CW and move the cursor to the right if not appropriately compensated (because handle rotation is mapped to cursor horizontal movement). If the participant produces a pronation torque to counter the perturbation torque, the cursor’s horizontal movement can be reduced. It is important to note that opposite wrist rotations, i.e., supination and pronation, were required for the left and right hands to compensate for the same perturbation direction.

Because it is well known that grip strength was affected by gender, handedness and aging, the scaling factor B was specified for each hand of each participant to account for such variability such that the perturbations pose similar levels of challenge to each individual’s grasp stability. This means that participants with different grip strengths would perform the task under similar relative loading conditions, minimizing strength-related confounds in the motor adaptation comparisons. Specifically, the perturbation torque at peak lifting velocity was approximately 200 N·mm for a participant with 120 N grip force at maximum voluntary contraction (MVC), and 140 N·mm for a participant with 60 N grip force at MVC. The scaling of perturbation magnitude was determined in our preliminary testing. Moreover, the perturbation magnitude was similar to our previous studies (20, 38), which was at a level that can be experienced during interactions with objects of daily life (e.g., turning a doorknob, holding a food plate, etc.).

Repetition exercise tasks

We designed two different exercise tasks targeting different aspects of hand control. The first one was force discrete (FD) repetition, which required producing ballistic isometric force pulses and was designed to focus on the predictive scaling of transient motor command. The second type was motion continuous (MC) repetition, which required producing continuous movement of the wrist guided by visual feedback and was designed to focus on the fine adjustment of motor commands in response to visual signals. In both types of exercise, participants were required to firmly grasp a handle that had the same shape as the handle used in the dynamic task, and to use the same grasp configuration as they did in the dynamic task. This is to ensure that participants do not take a mechanical advantage by strategically positioning their fingers differently during exercises of different directions.

For the FD exercise, participants were instructed to produce a torque (T_E) on the handle in the supination/pronation direction through their wrists and fingers as if they were trying to turn the handle. The handle was attached to a stationary structure via one force/torque (F/T) sensor Nano-25 (ATI Industrial Automation, Rochester Hills, MI) that measures the torque exerted on the handle (Fig. 2A). Therefore, the FD task is a pure isometric torque production task while participants applied wrist pronation/supination torque with minimal movements of the wrist and fingers. This torque is generated by combinations of finger forces but also activates wrist and arm muscles in isometric conditions for joint stabilization. For the MC exercises, the handle was attached to a rotary encoder that measures the rotation angle θ_E with resistance (Fig. 2B). The MC exercise is mostly a kinematic movement task. Here, participants performed wrist rotations with a light-weight handle and minimal friction during rotation. Therefore, only low finger forces were needed to rotate the handle. Both the torque and angle signals were acquired at 1000 Hz. In summary, while both tasks involve the same effectors (wrist and finger muscles), the FD task isolates force control without movement, and the MC task isolates movement control.

Figure 2:

Repetition exercise. A: Isometric torque production on a stationary handle. B: Movement repetition on a rotating handle. C: Task visual feedback showing a CCW torque being produced, corresponding to the direction of the torque or motion shown in Panels A and B, as an extending blue arc on a dashed circle. The open solid circle is the target. Note that this target requires wrist actions in the pronation direction D: Representative exercise schedules for one trial. 0 degree represents a target position at the top of the dashed circle that requires zero extension of the arc in Panel C.

To facilitate the exercises, we implemented a visual feedback system by mapping the signals from the F/T sensor and the encoder to the motion of an extending arc on a dashed circle, such that T_E and θ_E control the directional angular length of the arc (α) to track a target moving on the circle (Fig. 2C). The mapping between T_E and α was linearly scaled by the grip force MVC in a similar way as the dynamic task, such that participants with lower MVC would produce less torque to extend the arc for the same length. Again, such scaling was designed to minimize grip strength-related confounds. For example, a participant with 120 N grip force MVC needed to produce 240 N·mm torque to extend the arc to 100°. In contrast, the mapping between θ_E and α was 1:2 and invariant across participants. Therefore, 50° or wrist rotation extends the arc to 100°.

The temporal structures of the exercises were controlled by the movement of the target. In the FD exercise, the target followed a discrete trajectory and jumped to different locations with resting intervals of 1 s. Each position lasted 0.6 s to allow participants to generate torque pulses that extend the arc to acquire the target as fast and accurately as possible. The target’s color turns green if participants successfully hit the target (i.e., the peak of the torque pulse falls within ±10% error margin, indicated by the size of the target) within the time in which the target is displayed. The target circle returned to a default position (α = 0) during the resting interval, and the participants were instructed to relax. In the MC exercise, the target followed a smooth trajectory, and the participants were instructed to use the arc to track the target within ±10% error margin (if successful, the target turns green).

The magnitude of the target movement in both types of repetition exercises was governed by uni-directional sinusoidal function Asin(Bt+C)+D, where A, B, C, and D modulate the magnitude, frequency, phase, and bias of the waveform. By adjusting these parameters, the targets can appear to be either on the right or left side of the circle within one exercise trial to focus on motor actions in one direction, i.e., supination or pronation. The target position in the MC exercise directly followed the sinusoidal waveform, whereas the target positions in the FD exercise were sampled along the sinusoidal function (Fig. 2D). The overall duration of one repetition exercise trial was 40 seconds and included approximately 4 cycles of the waveform.

Experimental procedure

Before the start of experimental trials, we measured participants’ grip force during MVC. They were instructed to exert grip force with maximum effort on a F/T sensor with all fingers for three times using each hand, and each attempt lasted 3 s. The peak values of all attempts were averaged for each hand. Consistent with literature (39, 40), we found greater grip force MVC in YA and in right hands (YA: Right grip 103 ± 21.1 and Left grip: 93 ± 25.5, OA: Right grip 82 ± 21.9 and Left grip: 70 ± 16.3). These values were used to scale the strength of the perturbation in the dynamic manipulation task, as well as the magnitude of the FD exercise.

Participants were evenly and randomly assigned to one type of repetition exercise, i.e., FD or MC. They were then introduced to the dynamic manipulation task with 10–20 trials performed using their right hand in the Null dynamics, and to the repetition task with 2 trials to familiarize themselves with the tasks. Each participant performed 8 blocks of trials in the main experiment (Fig. 3), with less than 5 s inter-trial interval and 2 min inter-block rests. Specifically, participants performed three Null trials at the beginning of each block followed by a uni-directional repetition exercise (Supination- or Pronation-focused). After the exercise trial, participants performed another Null trial of the manipulation task, followed by 10 trials of the manipulation task with viscous force perturbation (CW or CCW) that needed to be compensated. To overcome the perturbation in the viscous field, each arm was required to produce the same or opposite direction of wrist action as it had been trained with during the preceding repetition trial. The presentation of perturbation dynamics was consistent within each block to allow participants to adapt. At the end of each block, the last four trials were Null conditions again to investigate the aftereffect and washout the adaptation to the perturbation dynamics. There were eight combinations of performing hand, direction of perturbation, and direction of exercise. For instance, the Block 7 required participants to use their left hand and they experienced a rightward perturbation during the dynamic manipulation task which supination movement of the hand was required to compensate the applied torque (T_C), they performed the opposite movement, pronation, during the exercise phase (T_E). Since there were only two perturbation directions, transfer of adaptation across blocks that share the same perturbation was a possibility. This confound was minimized by randomizing the order of the blocks, having a small number of perturbation trial in a block to prevent full adaptation, and including washout trials to promote de-adaptation.

Figure 3.

Trial order of one experimental block and different combinations of performing hand, direction of perturbation, and direction of exercise.

Data Analysis

Trials that were completed with more than 600 ms (too slow) and less than 400 ms (too fast) after the participants left the Start zone were considered as outliers. We excluded these trials (most were too slow) because they were typically incorrectly executed trials (e.g., cursor did not reach Target zone with a single lift), or the velocity-dependent perturbations in these trials were too different from the average amplitude. The total number of excluded trials was approximately 1% of all trials for both the YA and OA groups. We computed the feedforward motor error (Error) as a key behavioral variable to assess motor performance in our tasks across different task contexts. Error was defined as the maximum horizontal displacement of the cursor position from the center of the start sphere during the first 250 ms of each trial. This is consistent with past studies that implemented reaching tasks with similar (500 ms) movement duration (17, 38, 41–43). Note that we defined the error made in the direction that was the same as the perturbation direction to be positive. This was to allow direct comparison of performance error from different perturbation contexts.

To broadly estimate the rate of adaptation, we first averaged the Error for the same trial across eight blocks for each age group. Although participants were aware that a perturbation was going to be applied starting from Trial 5, they did not know the direction of the perturbation. Therefore, we did not include Trial 5 in the analysis to reduce the potential confound of surprise. We used a two-way mixed ANOVA with Age as a between-subject factor and Trial as a within-subject factor to examine the perturbation trials (Trial 6 – 14). This was followed by Helmert contrasts that compare each trial to the average of the following trials. Similarly, we also performed a two-way mixed ANOVA with Helmert contrasts to examine the NULL trials (Trial 16 – 18) after the perturbation ones. Based on the results of the contrasts, we defined three trial stages and averaged the participants’ Error within each stage to better quantify how motor adaptation was modulated by different factors in our experiments. The “Early” stage (Trials 6 – 9) started from the second perturbation trial after the repetition exercise. The “Late” stage includes last three perturbation trials (Trials 12 –14). Lastly, we defined the “Washout” stage which comprises the last three trials (Trials 16 – 18) in Null context over each block of the experiment (Fig.3).

The primary goal of this project was to determine the effect of Exercise type (FD or MC), which was used as a between-subject factor in a series of three-way mixed ANOVAs. These ANOVAs also included two within-subject factors in our task as shown in our block design: Direction of perturbation compensation T_C (Supination or Pronation), and the Similarity (Same or Opposite) between the directions of repetition exercise T_E and perturbation compensation T_C. These represent separate three-way ANOVAs for each age group (YA and OA), each hand (Right and Left) and different stages of trial (Early, Late, Washout). Post-hoc t-tests were used with Bonferroni corrections if statistically significant interactions were identified.

Our primary analysis revealed a significant effect of the Direction of perturbation compensation, i.e., there were differences in Error between perturbations that required supination and pronation to compensate by each hand. Furthermore, this effect was mostly independent of Exercise and Similarity. Therefore, we performed additional analysis to examine the extent to which this asymmetry was affected by Age, trial Stage, and performing Hand. Specifically, we defined a “Direction Asymmetry” metric that computes the difference between the averaged error of blocks that required perturbation compensation in the supination and pronation direction. This means that the same trial stages were averaged across different Similarity as defined above, and the participants from different exercise types were pooled. We utilized a two-way repeated ANOVA with within-subject factor Hand (Left versus Right) and Stage (Early versus Late). Additionally, we used Age as a covariate. Our statistical analyses were performed in MATLAB (“ranova”) and SPSS software (IBM, Armonk, NY).

RESULTS

The Null trials at the beginning of each blocks were performed with minimal errors in both YA and OA groups, suggesting that the participants encountered no difficulties executing the task in the absence of perturbation and no aftereffects from preceding blocks after de-adaptation and inter-block break. The first perturbation trial (Trial 5) induced larger perturbation errors since the participants could not predict the direction of the perturbation. For both YA and OA, improvements in performance were evident from the first to the last trial of the perturbation conditions within each block. Furthermore, we observed that OA exhibited overall lower performance than YA during perturbation trials. After averaging across all eight blocks and exercise groups for each trial from Trial 6 to Trial 14, a two-way mixed ANOVA revealed a significant main effect of Trial (F(8,464) = 35.99, p < 0.001, ƞ_P ² = 0.383) and a significant effect of age (F(1,58) = 7.56, p = 0.008, ƞ_P ² = 0.115). These results confirmed our observations (Fig. 4). Additionally, Helmert contrasts were subsequently employed. The contrast indicated a significant difference in the performance errors between Trial 6, 7, 8, 10 and the subsequent trials (p < 0.002), while results were not significant for Trial 9, 11, 12, 13. This finding suggests a relatively rapid adaptation before the fifth perturbation trial, with learning appearing to be relatively slow after the sixth perturbation. After the perturbation was removed, the performance error reverted direction, showing aftereffects that underwent trial-to-trial de-adaptation in both YA and OA groups. This was confirmed by a two-way mixed ANOVA showing only a significant main effect of Trial (F(2,116) = 34.82, p < 0.001, ƞ_P ² = 0.353). Additionally, Hemert contrasts showed that both Trial 16 and 17 were significantly different from the subsequent trials (p < 0.003). Based on these results, we defined three trial stages ‘Early’, ‘Late’, and ‘Washout’ for the following analyses.

Figure 4.

Performance error for YA and OA averaged across eight blocks and exercise groups (mean ± S.D.).

Effects of exercise type and compensatory torque direction on task performance in Young Adults.

We first did in-depth analysis within each trial stage of the YA groups to examine the effect of Exercise type, Direction of T_C, and the Similarity between the directions of T_E and T_C. For the right hand of YA, we found that Error was smaller for participants with FD exercise than those with MC exercise only in the Late stage. Additionally, we observed that perturbation that required supination T_C induced less error than the opposite perturbation in both Early and Late stage. Lastly, the similarity between the directions of T_E and T_C did not make any differences in task performance. These findings were confirmed by 3-way ANOVAs. In the Early stage, there was only a significant effect of Direction (F(1,28) = 4.44, p = 0.044, ƞ_P ² = 0.137; Fig. 5A). In the Late stage, there was a significant effect of Direction (F(1,28) = 7.76, p = 0.009, ƞ_P ² = 0.217) and Exercise (F(1,28) = 4.83, p = 0.036, ƞ_P ² = 0.147; Fig. 5B). In the Washout stage, no significant effects were found (Fig. 5C).

Figure 5.

Performance error in young adult groups (mean ± S.D.). A-C: Error of the right hand for different exercise groups over 3 stages. D-F: Error of the left hand for different exercise groups over 3 stages. Data from two exercises T_E directions were averaged if they shared the same compensatory torque T_E direction. Two-colored squares and asterisks denote significant main effect of Direction (p < 0.05). Plus (+) denotes significant main effect of Exercise.

For the left hand of YA, participants with FD and MC exercises performed similarly in all stages. Furthermore, we observed perturbation that required supination T_C induced less error than the opposite perturbation in the perturbation trials, and the similarity between the directions of T_E and T_C did not make any difference in most cases. These findings were confirmed by 3-way ANOVAs. In the Early stage, there was only a significant three-way interaction (F(1,28) = 4.73, p = 0.038, ƞ_P ² = 0.145). Post-hoc t-tests revealed significant differences between two perturbation directions in participants with FD exercise (p < 0.002), as well as in participants with MC exercise when T_E and T_C were the same direction (p < 0.001). In the Late stage, there was only a significant effect of Direction (F(1,28) = 41.45, p < 0.001, ƞ_P ² = 0.597; Fig. 5E). In the Washout stage, there was also a significant effect of Direction (F(1,28) = 8.62, p = 0.007, ƞ_P ² = 0.235; Fig. 5F).

Effects of exercise type and compensatory torque direction on task performance in Older Adults.

In the examination of OA within each trial stage, a detailed analysis was conducted to investigate the effect of Exercise type, Direction of T_C, and the Similarity between the directions of T_E and T_C, mirroring the approach taken in the analysis of YA. For the right hand, we observed that participants with FD and MC exercises performed at similar levels across all stages. Additionally, the direction of T_C and the similarity between the T_E and T_C did not make any difference in task performance. Lastly, we found that the perturbation that required supination T_C induced less error than the opposite perturbation only for the Washout stage. These findings were supported by a 3-way ANOVA in the Washout stage indicating a significant main effect of Direction (F(1,28) = 6.50, p = 0.016, ƞ_P ² = 0.187; Fig. 6C), and we did not find significant effects of any factor in the Early and Late stages (Fig. 6A and 6B).

Figure 6.

Performance error in older adult groups (mean ± S.D.). A-C: Error of the right hand for different exercise groups over 3 stages. D-F: Error of the left hand for different exercise groups over 3 stages. Data from two exercises T_E directions were averaged if they shared the same compensatory torque T_E direction. Two-colored squares and asterisks denote significant main effect of Direction (p < 0.05).

For the left hand of OA, participants with FD and MC exercises performed similarly and the similarity between the directions of T_E and T_C did not make any differences in task performance across all stages. Additionally, we observed that the perturbation that required supination T_C induced less error than the opposite perturbation only for the Late stage. This finding was confirmed by a 3-way ANOVA showing a significant main effect of Direction during late stage (F(1,28) = 19.34, p < 0.001, ƞ_P ² = 0.409; Fig. 6E). ANOVAs for the Early and Washout stages did not reveal any significant effects (Fig. 6D and 6F).

Effects of hand and age on the supination/pronation performance asymmetry during perturbation stages.

Our analyses unexpectedly unveiled a significant effect of the Direction of perturbation compensation in some conditions and subject groups, which indicates asymmetrical capabilities to adapt to the perturbation between supination and pronation directions. A closer observation of such asymmetries suggested that they differed between the two hands and between age groups. Therefore, we conducted additional analysis in the Early and Late stages by defining the Direction asymmetry. We found that the left hand exhibited a greater asymmetry than the right hand did, and OA showed less asymmetry than YA (Fig. 7). This was supported by a two-way ANOVA with Age as a covariate, demonstrating a main effect of Hand (F(1,58) = 5.45, p = 0.023, ƞ_P ² = 0.263) and a main effect of Age (F(1,58) = 4.71, p = 0.034, ƞ_P ² = 0.214), with no interaction between factors. Furthermore, we performed a correlation analysis between the right and left hands after averaging across the Early and Late stages, which showed a significant correlation among all 60 participants (Pearson’s r = 0.382, p = 0.003).

Figure 7:

Direction asymmetry for the right and left hand among all participants. The red dotted lines denote results of linear regression.

DISCUSSION

We observed that YA exhibited greater adaptation than OA across perturbation trials. Notably, in YA right hands, participants who performed FD exercise showed smaller errors in the late adaptation stage compared to those who performed MC exercise, indicating FD exercise was more effective in facilitating the motor adaptation. In contrast, exercise types did not significantly differ in their effect on adaptation in OA. Additionally, across both age groups, perturbations requiring supination torque compensation generally resulted in lower errors than those requiring pronation, revealing a directional asymmetry that was more pronounced in YA and in the left hand.

Effect of repetition type in the right hands of young adults

The well-established framework of hemispheric lateralization of motor control argued that the right hand and left hand may adapt to mechanical perturbations differently (44). During fast reaching, the dominant arm adapts to force field perturbations with a greater degree of reliance on developing accurate predictive motor commands than the non-dominant arm (45). Note that the adaptation of predictive control in object manipulation can be driven by both error-based updates of internal models and alternative mechanisms that are not dependent on error signals such as use-dependent adaptation (12). In our experiments, the repetition exercise did not provide any error feedback that was relevant to the main manipulation task. Therefore, it is impossible for the repetition exercise to directly contribute to the error-based adaptation of the internal model during the manipulation task. Nevertheless, use-dependent adaptation during the repetition of force pulses can still produce biases in the predictive motor commands of finger forces. It was theorized that the use-dependent adaptation may originate from the primary motor cortex (M1) (10). Given many differences observed between the dominant and non-dominant M1 (46), it is possible that the lack of repetition type effect in the left hand was due to a weaker ability to produce the use-dependent adaptation, or due to a different interaction between these parallel adaptation mechanisms in the left hand.

Our results also showed that the movement repetition did not produce the same level of use-dependent adaptation as force repetition, although it was performed by the same end-effector and muscles. The neural control of force and movement in human upper limbs remains to be an ongoing debate. One hypothesis proposes that the central nervous system (CNS) has one position/impedance controller for both types of motor tasks. When the limb exerts forces against a surface, it controls a ‘reference’ position (i.e., equilibrium point) inside the surface and regulates the viscoelastic muscle behavior. A contact force would emerge from the mismatch between the actual and reference position, without directly encoding it in the neural signal (47–49). Alternatively, it was hypothesized that the CNS implements two distinct controllers for force and movements and switches between the two when contact is made or released (50–52). Regardless of where one may stand on this debate, our results suggest that the type and format of motor repetition are important factors to consider for maximizing the effect on subsequent motor tasks.

A surprising and important result is that we did not find an effect of the exercise direction. That is, whether the direction of repetition was the same as the compensatory action did not result in different effects of the repetition. At first glance, this may not align with the idea of repetition-induced bias, which was expected to be directional. However, two recent studies showed that the motor control of multi-finger dexterous manipulation may consist of two components that are controlled and adapt differently: a directional manipulation force and an internal force without a direction (38, 53). Specifically, finger forces from multiple fingers form a redundant system and these forces can be mathematically decomposed based on their effects on the object. The manipulation force component (e.g., a torque) can directly move an object or counter a perturbation in a specific direction. For a manipulation task like ours, the manipulation force mostly comes from the tangential forces from the conventional view of finger force control. In contrast, the internal force component is the portion of finger forces that act against each other through the object and do not directly generate movement, but it can resist external disturbances from all directions by acting as a group of springs. The internal force component mostly involves normal forces in the present study. These two force components are not completely independent. While internal force can be altered without changing the manipulation force, an increase in manipulation force likely requires an increase in internal force given the physical relation between normal and tangential forces (i.e., tangential forces are proportionally supported by normal forces). Therefore, we speculate that the use-dependent adaptation during the force repetition may only operate on the internal force control, generating a bias in each finger to facilitate a stronger internal force magnitude. In other words, the directional FD repetition, which was originally designed to focus on the repetition of the manipulation component of the finger forces, also engages the repetition of non-directional internal force. However, unlike manipulation force, the internal force is independent of the exercise direction, i.e., both CW and CCW exercises would require increasing the internal forces in similar ways. In our previous work, we showed that the internal force can facilitate the adaptation to torque perturbations in a motor interference paradigm, where the internal force component can be modulated by context switches, thereby reducing the effect of contextual interference in the manipulation force component (38). Based on this theory, YA participants may be able to ‘squeeze harder’ after the FD exercise in either direction, therefore improving the internal force facilitated error reduction during perturbation. Future studies that measure individual finger forces are needed to examine this speculation. Interestingly, a recent study using arm reaching tasks showed that pre-exercise a co-contraction strategy that increases arm stiffness, and explicit instruction of maintaining high stiffness can accelerate adaptation in a force field (54). A co-contraction strategy in reaching functions similarly as the internal force component in manipulation, both resist perturbations independent of directions. This is consistent with our results. However, directly comparing our results to Heald et al. (2018) must be taken with caution, because our protocol did not explicitly train and instruct the participants to use a strategy that uses high internal force or co-contraction. Rather, any benefits that may come from the repetition of the internal forces were likely a direct result of implicit use-dependent adaptation.

Training type did not modulate adaptation in older adults

Our results on overall adaptation showed that YA outperformed OA during exposure to perturbation, but both age groups exhibited similar aftereffects after the perturbation was removed (Fig. 4). This is consistent with existing studies of motor adaptation in OA. As we reviewed in the Introduction, it was proposed that motor adaptation during perturbation exposure, but not trials after exposure, engages both implicit and explicit adaptation mechanisms (King et al., 2013). OA’s worse performance during exposure can therefore be attributed to a deficit in explicit adaptation with cognitive strategies. In contrast, the post-exposure phase is mostly related to implicit sensorimotor recalibration, which does not change significantly with age. This explanation aligns with the underlying age-related changes in neural structures. The cortico-striatal network is known to play a crucial role in explicit motor learning such as motor sequence learning (55, 56), whereas the cortico-cerebellar network is essential for implicit motor adaptation (6). It has been shown that the cerebellum may be relatively better preserved in OA than the striatum (57), supporting the aforementioned age-related changes in motor adaptation.

Our results revealed additional differences between YA and OA groups, as the two repetition exercise types did not result in different adaptation behavior in the OA group. Aging is shown to be associated with decreased asymmetry in some unimanual movements. For instance, movement trajectories and accuracies of OA were found to be more symmetric than YA in fast-reaching and slow-tracing tasks (58, 59). One explanation underlying the more symmetric motor behavior in OA could be that each hand may rely on both hemispheres to a greater degree in comparison to YA (60). Based on this view, it is possible that OA may use less predictive control and more feedback-driven mechanisms with their right hands in our task, thus diminishing the benefit of force repetition on predictive force control in the right hands. This is consistent with the literature that shows a decline in the capability of predictive finger force scaling of the dominant hands in OA (61).

Another explanation is that older adults tended to use higher grip force to maintain a higher safety margin during grasping (61–63), similar to the higher level of muscle co-activation found in OA performing reaching movements (64) and walking (65). Such an increase in grip forces may represent a learned compensation strategy for the age-related increase in skin slipperiness and reduction in tactile information (33). As discussed above, an elevated level of grip force can lead to high internal force and the use-dependent adaptation may have a diminished effect on high force levels (66), thus making the effect of FD exercise less effective in OA. Furthermore, it is well established that motor behavior can become more stereotyped as people age, and older adults have a decreased ability to use flexible explicit strategies (67, 68). It is possible that the reduced motor flexibility leads older adults to rely on such habitual, over-learned force patterns in the control of dexterous manipulation. Therefore, in the present study, the reliance on rigid stereotyped force strategies in older adults may impede the short-term effect of use-dependent adaptation experienced in the preceding repetition exercise. However, future studies are needed to test this theory since we did not measure grip forces in our experimental tasks.

Lastly, it is possible that our FD exercise protocol was not optimal to generate adequate use-dependent adaptation in OA. The torques participants generated in the exercise did not strictly match those required during the main motor adaptation task. The FD task included multiple target levels (Fig. 2D), requiring varying amounts of torque across repetitions, some of which approached the required compensatory torque levels for the motor adaptation task, while others did not. Importantly, participants were not consistently asked to produce maximal torque during FD repetitions. From the perspective of use-dependent adaptation, closer matching between the exercise and test tasks (in terms of magnitude and direction of motor output) would likely enhance the specificity and magnitude of exercise effects (e.g., Diedrichsen et al., 2010; Verstynen & Sabes, 2011). However, in our design, we intentionally used a range of torque levels to maintain task variety and avoid fatigue. As a result, our FD repetition likely fostered an average or intermediate level of use-dependent adaptation, rather than precise adaptation to the exact demands of the dynamic manipulation task. This framework may also partially account for the lack of directional effects of exercise in YA. Nevertheless, it is important to point out that, although exact matching exercise dynamics to the generalization task may promote stronger use-dependent adaptation that benefits the adaptation of a specific task, a wider range of repetition can be more practical from a clinical perspective to avoid fatigue, maintain engagement, and prevent over-adaptation.

Differences in performance between pronation and supination directions

We found that participants generally performed better when a supination torque was needed to compensate for the perturbation than a pronation torque. This is consistent with the previous object lifting studies showing the anticipatory torque generation was more accurate when the objects had leftward weight distribution that required supination of the right hand (69, 70). Nevertheless, this phenomenon has not been explained. Our study only scaled the strength of the perturbation and exercise by the MVC of the grip force, and we did not systematically scale these task parameters with supination and pronation torque production capability. Therefore, one potential explanation for the overall superior performance in the supination direction is that there may be asymmetries in the torque production capability between supination and pronation. Previous measurements of torque production showed inconsistent results, with some reporting a stronger pronation torque (71) and others reporting equal or a stronger supination torque (72, 73). Moreover, these measurements were taken with participants exerting torques using a power grasp configuration. In contrast, our task relied on fingertips to produce the compensatory torque. We decided to take additional measurements to assess the torque production capabilities using a grasp configuration similar to the one that was required to perform the dynamic manipulation and repetition exercise. A subgroup of participants (Table 1) was evaluated. With a three-way ANOVA (Age x Direction x Hand), we found only a significant main effect of Hand (F(1,20) = 9.219, p = 0.007). This result indicates that the torque production capability is unlikely to be the main factor that leads to asymmetrical performance in our dynamic manipulation task, because supination and pronation strength were similar in our study.

Table 1.

Average maximum voluntary wrist rotation torque (mean ± S.D) for each hand.

Group	Left Hand (Nm)		Right Hand (Nm)
	Supination	Pronation	Pronation	Supination
YA (n = 10, 3M)	2.62 ± 0.79	2.67 ± 0.88	2.80 ± 0.81	2.91 ± 0.86
OA (n = 8, 7M)	1.94 ± 0.65	1.93 ± 0.56	2.04 ± 0.44	2.16 ± 0.60

An alternative explanation for the directional asymmetry observed in this study is the differences in the inter-finger coupling during motor control. It is important to note that supination torque production primarily uses the ring and little fingers, whereas pronation torque production primarily uses the index and middle fingers (74). There has been strong evidence that the degree of finger individualization is different between fingers. Specifically, the ring and little fingers were shown to have a higher level of coupling, i.e., ‘enslaving effect’, than the index and middle fingers during isometric force production (75). Additionally, it was found that the limited independence of ring and little fingers in movement tasks is likely constrained by neuromotor control instead of biomechanical factors (76). Therefore, it is plausible that the higher degree of synergistic control between the ring and little fingers is advantageous for generating compensatory torque in the supination direction. Interestingly, the strength of inter-finger coupling is affected by aging. For instance, a higher level of enslaving effect was found in older adults in both static and dynamic force production (77). Furthermore, it was revealed that the movement independence of the index finger was reduced by aging, likely due to more co-activation between finger-specific muscle regions (78). Such an increased enslaving effect in older adults is in agreement with a decreased supination advantage over pronation in older adults of our study. However, the movement independence and the enslaving effect during isometric force production were shown to be similar between the dominant and non-dominant hands (79, 80). This indicates that the inter-finger coupling may not fully explain the directional asymmetry in our task, as we found the left hand has a stronger supination advantage. Further experimental studies are needed to uncover the underlying mechanisms for the asymmetry between supination and pronation.

LIMITATIONS

This study has some limitations that should be acknowledged. First, we did not record individual finger forces during the dynamic manipulation task. This was a design choice we made to limit the total number of sensors needed, to minimize the size and weight of the handles, and to enable a more natural grasp on the handle. As a result, we were unable to directly quantify the effect of motor repetition on force control at the individual finger level. We will include finger force sensors in future studies to investigate our speculation about the effect of force repetition on internal force control.

Second, the distribution of male and female participants was not balanced, particularly within the OA group. This sex imbalance was mainly caused by enrolling participants on a rolling basis, and there can be a sex-related bias within the local older adult population in responding to recruitment. The most notable sex-related difference is grip strength, and we tried to minimize the impact of grip strength difference by scaling perturbation and FD exercise forces based on individual MVC. There has not been a well-known sex-related difference in motor adaptation. For example, past studies did not find a difference in adaptation rates in visuomotor adaptation paradigms between females and males (68, 81, 82). Additionally, existing studies did not show sex-related differences in the symmetry of supination/pronation strength. Therefore, we believe that sex imbalance is not a major confounder in the present study. However, we cannot completely rule out lesser-known sex differences in our task, which has a novel and unique experimental design based on force repetition exercises and force field perturbations.

Lastly, our sample size was relatively small. This sample size cannot detect small effects between the two exercise groups. Additionally, our YA group used a relatively wide range of age eligibility (18 – 45 years). This caused our average age of the YA group being older than some other studies (two individuals in YA group were > 40 years and most were < 30 years). Although we don’t believe that there are significant age-related changes in motor function before 45 years, it is possible that we had more variance in participant characteristics. Moreover, our OA group was on the younger end of the OA spectrum (55 – 71 years), and they were likely generally more active than the average OA population. Therefore, our small sample size may limit the generalizability of the present study.

CONCLUSIONS

This study investigated how different forms of simple motor repetition, force-based or movement-based, influence subsequent adaptation to dynamic perturbations during a dexterous manipulation task, and how these effects vary with age and hand dominance. Our findings demonstrate that in young adults, force repetition facilitated better late-stage adaptation in the dominant hand compared to movement repetition, suggesting a role for use-dependent mechanisms in priming predictive motor control. However, this benefit was absent in older adults, highlighting age-related changes in adaptation processes and potential shifts in reliance on predictive versus feedback control strategies. Additionally, we observed a consistent asymmetry in adaptation performance between pronation and supination directions, with supination showing superior performance. This directional effect varied by age and hand, pointing to complex interactions between neuromuscular control, finger coordination patterns, and aging. Overall, these findings underscore the importance of tailoring repetition-based exercise approaches to individual motor profiles and age-related capabilities. They provide critical insights for developing more targeted rehabilitation protocols aimed at enhancing motor learning and adaptability in both young and older populations.

DATA AVAILABILITY

Data will be made available upon reasonable request.