A Computer Interface Controlled by Upper Limb Muscles: Effects of a Two Weeks Training on Younger and Older Adults

As the population worldwide ages, there is a growing need for assistive technology and effective human-machine interfaces to address the wider range of motor disabilities that older adults may experience. Motor disabilities can make it difficult for individuals to perform basic daily tasks, such as getting dressed, preparing meals, or using a computer. The goal of this study was to investigate the effect of two weeks of training with a myoelectric computer interface (MCI) on motor functions in younger and older adults. Twenty people were recruited in the study: thirteen younger (range: 22–35 years old) and seven older (range: 61–78 years old) adults. Participants completed six training sessions of about 2 hours each, during which the activity of right and left biceps and trapezius were mapped into a control signal for the cursor of a computer. Results highlighted significant improvements in cursor control, and therefore in muscle coordination, in both groups. All participants with training became faster and more accurate, although people in different age range learned with a different dynamic. Results of the questionnaire on system usability and quality highlighted a general consensus about easiness of use and intuitiveness. These findings suggest that the proposed MCI training can be a powerful tool in the framework of assistive technologies for both younger and older adults. Further research is needed to determine the optimal duration and intensity of MCI training for different age groups and to investigate long-term effects of training on physical and cognitive function.


I. INTRODUCTION
A CCORDING to United Nations and the World Health Organization the global population over 60 years is projected to nearly double from about 900 million to almost 2 billion people by 2050 [1].As the population ages, an increasing number of older adults are experiencing physical and cognitive impairments that can lead to very severe motor disabilities like the ones following a stroke or a cervical spinal cord injury that not that long time ago were mostly involving younger people than nowadays.Such disabilities can make difficult for them to access and use technologies or other helping aids like powered wheelchairs [2], [3].Indeed, technology can assist in simple situations like healthy aging and physical activity for older adults [4], [5], [6], [7].But more importantly technology can greatly benefit older adults with impairments by helping them stay connected with their loved ones, access important information, and participate in social and leisure activities.Therefore, access to technology is crucial for people with an impairment because it can help to reduce social isolation, improve mental health, and increase independence [8], [9], [10].
In this context, it is important to explore the various use of technology to improve the quality of life for older adults with impairments, and ensure that these technologies are accessible and user-friendly for this population.In recent years, a range of human-machine interfaces (HMI) have been developed specifically to support and assist people with impairments, such as tangible HMI, like specialized keyboards or mouse, and touchscreens with large buttons, and not tangible HMI like voice-activated assistants [11].Unfortunately, in some cases people have unique needs because they might be no longer able to move any or most body parts or have difficulty speaking, therefore alternative ways to control assistive devices are needed and have to be studied in research going in the direction of nontangible HMI.A general approach in this context is to use signals from the brain or from other part of the body, like eyes [12], tongue [13] or shoulder and arm movements [14] or muscles activities [15], [16], to control assistive and/or rehabilitative devices.The conversion of brain signals into commands for devices forms the central focus of research on brain-machine interfaces (BMIs); in contrast body-machine interfaces (BoMIs), do not bypass the final motor pathway and capitalize the user's existing motor abilities, holding the potential to enhance these skills through consistent utilization.BoMI is a broad term that indicates interfaces translating signals from the human body into control command for a device like a prosthesis, a drone, a powered wheelchair or a computer.It involves recording and interpreting signals or data from the human body, such as muscle contractions, kinematics or other physiological responses to generate control commands.MCI is a specific type of BoMI that focuses on the use of muscle activity (EMG) for controlling a computer.MCI have been proven to have the potential to improve independence while also enhancing recovery trough personalized interventions [17], [18].An MCI detects the electrical activity from muscles (EMG) to control computers and other external devices.By harnessing the neurological signals needed for movements, MCIs can provide intuitive and personalized control of assistive devices, virtual reality systems, and smart environments.BoMIs based on movement might be easier to control but have, however, a big limitation of not offering the capability to selectively engage specific muscles in the operation.In contrast, MCIs directly map the activity of targeted muscles into control input for the device, but they face challenges such as recording instabilities, sensitivity to electrodes placement and increased uncertainty in the control due to the lower signal-to-noise ratio.However, recent advances in electromyography sensing, signal processing, and machine learning have increased the reliability, dexterity, and intuitiveness of myoelectric control [18], [19], [20], [21], [22] and there is also the possibility to use hybrid interfaces that allows combining motion and muscle signals [19].Moreover, MCI can be integrated with virtual reality or augmented reality systems to detect user movements and provide interactive feedback within immersive or augmented virtual environments, enhancing not only independence but also motor learning and movement coordination [18], [23], [24].
The motor system, which is responsible for controlling our movements, undergoes significant changes with age [25], [26], [27].These changes can result in differences in motor control, reaction time, and motor coordination between young and old individuals [28], [29].In the context of developing a HMI and more in particular an MCI, it is important to understand these age-related changes in the motor system because they can impact the design and effectiveness of the interface [27], [29], [30], [31].For example, if the interface requires precise motor control, older individuals may have difficulty using it due to declines in fine motor skills.Similarly, if the interface requires quick reaction times, older individuals may not be as effective as younger individuals due to declines in processing speed.By conducting a study that includes both younger and older adults, we can identify these age-related differences in the motor system and use this information to design an MCI that is effective for both age groups.This can lead to a more inclusive and accessible interface that can be used by people of all ages and abilities.
This paper developed a simple and intuitive myoelectric computer interface as assistive technology for controlling the cursor of a computer and examined the use of such MCI by two groups of people in different age-ranges: younger and older adults.The MCI transformed the activation of four main muscles of the upper limbs into the control signal for moving the cursor of a computer while subjects had to complete a series of engaging games.The training with the MCI lasted two weeks and the main goal was to identify the effects of the long training on the performance of the two groups and in relation to the age.In particular we wanted to test the following hypothesis: i) the two groups (younger and older adults) would have different level of performance; ii) all the subjects would learn how to use the MCI across different days with higher improvements at the beginning of the training period; iii) age would have different effects on the learning process.In the future the proposed MCI can also be used as a tool for promoting continuous engagement of elderly into physical activity exercises targeting specific muscles.

A. Participants
Twenty subjects without prior history of neuromotor or musculoskeletal disorders were recruited in the study.13 subjects were part of the younger adults group (age: 26.5±3.5 STD yo) and seven subjects of the older adults group (age: 66.1±5.8STD yo).
Inclusion criteria for being part of the younger adults group was to be 35 years old or younger while for being included in the older adults group was to be 60 years old or older.
The study was conducted at the University of Genoa in accordance with the ethical standards laid down in the Declaration of Helsinki (2013 revision) for the protection of research subjects, all the study procedures were approved by the Institutional Review Board (Comitato Etico DIBRIS, registry number: 009/2020), and each participant gave written informed consent prior to their participation in the study.

B. Experimental Apparatus
The MCI used in this project recorded the surface electromyographic (sEMG) activity of four muscle of the upper body, in particular right and left long head of the biceps and right and left descending trapezii, and transformed it into a control signal for the cursor of a computer (Fig. 1).EMG signals were recorded using the Zerowire system (Cometa Systems, formerly known as Aurion) with a sampling frequency of 1.5 KHz.The raw EMG signals were first bandpass-filtered (5nd order FIR, fc = 40 − 450 Hz), then rectified and lowpass-filtered (3nd order FIR, fc = 2 Hz) to obtain the EMG envelope [16].The envelope was normalized by the maximum voluntary contraction (MVC) recorded at the beginning of the session during the calibration phase.In this phase the subjects were asked to contract for 5 times within a time interval of 20 seconds the selected muscle at their maximum capacity without relying on any surface support, then for each The envelope of each muscle normalized by the MVC ('m') was then used to move the cursor in a specific direction in the computer screen.
More precisely, considering the center of the screen as the origin of a cartesian plane (x-y), the map between the subject's muscular activity and the cursor on the screen was as follow: where xi and yi, xi-1 and yi-1 denote the position of the cursor along the horizontal an the vertical axes at time instant i and at the previous time instant i-1; mNi indicates the normalized envelope values of the N muscle (N=1. . .4) at the ith instant and k is a gain equal to 0.003.Thus the cursor's x and y coordinates at the instant ith with respect the cursor's position at the instant (i-1)th were incremented proportionally to the differences between the two pairs of muscle envelopes, m1-m2 and m3-m4 at the i-instant.Specifically -The signal from the right biceps (m1) controls the movements along the positive x axis (rightward) -The signal from the left biceps (m2) controls the movement along the negative x axis (leftward) -The signal from the left trapezius (m3) controls the movement along the positive y axis (upward) -The signal from the right trapezius (m4) controls the movement along the negative y axis (downward).
Thus, the signals coming from all four muscles involved allowed to extend the workspace to the whole screen.Notice that 'm' indicates a signal that is always positive so the positive or negative motion is determined by the sign preceding the signal in the equations: + positive motion along the designed axis, -negative.Therefore if a subject activates muscle 1 and muscle 2 (or muscle 3 and 4) equally, both muscles will contribute equally to the horizontal movement, but in opposite directions.In such case, the effects of the two muscles would cancel each other out, resulting in a stationary cursor.

C. Training Paradigm
The experimental protocol consisted in six training sessions: three times a week for two weeks, see fig. 2. Each session lasted maximum 2 hours and was composed of a calibration Fig. 2. Training paradigm organized in six days over two weeks, three times a week.Each session lasted about two hours and was organized in a calibration phase (C) followed by a test block (T), during which subjects were performing a reaching task for 5 minutes and then a typing task, then a 30 minutes free PC use (F) where subjects were playing videogames and free writing of emails or text files.After that, they repeated the test block with the same modality of the first test block.At the end of the two weeks training they were also asked to fill a questionnaire to evaluate the MCI.
block, a test block before the block of free PC use and one test block after it.During each session, participants seated in front of a 15.6" computer screen, placed at a distance of 0.7 m and went through a calibration block, then a test block followed by a block of free PC use and the session ended with a second test block.In the calibration block the MVC of the four upper body muscles were recorded to create the map from the body to the cursor space as described in "Experimental apparatus".The test block consisted in a Reaching task and a Typing task.Working with older adults, thus easily fatigable subjects, we decided to adopt a fixed time design for the reaching and typing tasks.In the first task subjects have 5 minutes to reach the highest number of targets that were appearing in different positions in the screen with a pseudo random order.The subject had 15 seconds to move the cursor, a circle of 0.5 cm radius, from one target (radius 1 cm) to another one otherwise the target disappeared and a new one would appear.A target was considered reached when the subject was holding the cursor inside the target for 0.5 seconds.When the cursor was correctly positioned inside the target, the target changed color from pink to green.In the typing task a virtual QWERTY keyboard would appear in the upper half of the screen and a notepad page would open in the lower part of the screen.The dwelling time to select a key for display was set at 0.5 seconds.Subjects had to type the Italian pangram sentence "ma la volpe col suo balzo ha raggiunto il quieto fido" in maximum 30 minutes.This task required the user to stabilize the cursor over each key for a successful character display.Failure to stabilize the cursor resulted either in the inability to generate text or in the production of erroneous text (i.e., the cursor rested over the wrong key), requiring corrective action.Participants were instructed to correct for typing errors.The free PC use lasted 30 minutes and it was organized as 10 minutes of games playing time, 10 minutes of free text writing and other 10 minutes of videogames playing time.At the end of the training period, subjects had to fill a questionnaire, translated and readapted from the original Post-Study System Usability Questionnaire (PSSUQ, [32]), to evaluate MCI usability and interface quality.The questionnaire was presenting twelve sentences, seven about system usefulness and five about interface quality, that the subjects had to score from 0 to 7 with 0 'I fully disagree' and 7 'I fully agree with the sentence'.For more details on the questions of PSSUQ see Supplementary Material.

D. Data Analysis
We evaluated the performance of each subject during the two tasks of the test blocks.In particular for the reaching task we computed the following parameters: -Target Hits, number of targets that the subject correctly reached within the predetermined duration of the reaching task i.e. 5 minutes.The number of target hits can be less or equal to the number of total targets because if a target was not reached within 15 seconds from its appearance, it would disappear, and a new target would appear for the subject to reach.
-Target Missed, number of targets the subject was not able to reach within the 15-seconds time window given to reach each target.
-Normalized Target Capture Time (NTCT), the time necessary to reach a target after its appearance normalized by the path length, i.e., the time between the appearance and disappearance of a target minus the time of maintenance of the cursor on the target (0.5 s), divided by the distance between the initial and final position of the cursor in those corresponding time instants.
In the typing task we computed: -Task duration (TD), the time the subject took to type the full pangram (TD ≤ 30 minutes); -Correct letters (CL), the number of correct letters typed by the subject (ideally CL=54); -Number of correct letters over task duration, as the ratio between the two previous parameters that is giving an idea of the "speed" in typing; -Extra Actions, as the number of keys pressed to correct for typing errors.
We consider target hits and number of correct letters over task duration as the primary outcomes of respectively the reaching and typing tasks.To check whether there is a relationship between age and tasks performance in the group of older adults we run a polynomial fitting using a least squared approach on the primary outcomes metrics of the reaching and typing tasks of the older adults group computed in the first test block of day 1 and in the last test block of day 6.

E. Statistical Analysis
We expect that younger and older adults will behave differently but at the same time we expect that they all learn to use the interface across the six training sessions and within the same session.Therefore, to verify that, we run a repeated measure ANOVA with one between subjects factor: age group, and two within subjects factors: block (beginning vs end session) and day (day1 vs day6) on all the parameters computed from the reaching and typing tasks.We considered as primary outcomes the target hits for the reaching task and number of correct letters over task duration for the typing task.Additionally, to investigate the presence of different effects to the learning dynamics due to age we also check for significant interactions between age group and session, and between age group and training.
The post hoc analysis was performed by using the Tukey test.We corrected for multiple comparison by using Bonferroni.
To check for group differences in the results of the questionnaire we run a Wilcoxon rank sum test on each of the twelve questions of the PSSUQ.
Prior to any statistical tests, normality was checked with the Shapiro-Wilk test.All parameters resulted normally distributed.Statistical significance was set at p<0.05.The statistical analyses were performed within Jamovi environment (Jamovi software 0.9.2.8).
To assess the goodness of the polynomial fitting for the primary outcomes metrics of reaching and typing tasks of the older adults group we reported the R 2 and the p-value of the fitting.

III. RESULTS
With practice, all participants improved their control skills.Indeed in the reaching task both groups from the first to the last training day and also within the same day significantly increased the number of targets acquired and the number of targets presented in the 5 minutes duration of the reaching task, Fig. 3(a).This is also confirmed by a significant decrease of the number of targets missed, Fig. 3(b).Moreover they became faster in reaching the targets as shown by the NTCT, Fig 3(c).In all these performance metrics there was also a significant interaction between factors 'block' and 'day' due to a faster learning within session at the beginning of the training (day 1) with respect to the end (day 6).For details on statistical results see Table I.Indeed, when looking at the post-hoc analysis for all reaching metrics there is a significant difference between the beginning and last block of day 1 (p<0.001)that is then lost when comparing the blocks of the last training day.As expected there was also an effect of age, see Table II.Interestingly we also found a significant interaction 'age' x 'day' and 'age' x 'block' for the target missed meaning that younger and older adults groups were learning to control and use the MCI with different learning dynamics within the same block, starting from worst performance older adults improved more, and also in different days.Indeed, while younger subjects reached a stable performance at day 6 older subjects still showed a small trend of improvement within session in this last day of training.
A similar trend of improvement is evident also in the typing task, see Fig 4 .Both groups took significantly less time to complete the task (Fig. 4(b)), in particular the older group that was not able to write the full sentence since day one increased also the number of correct letters typed within the available typing time (Fig. 4(c)).Both groups with training became faster, as highlighted by the correct letters over task duration parameter (Fig. 4(c)), and decreased the number of extra actions (Fig. 4(d)).For details on statistical results see Table I.Also in this case the improvement was significant not only across days, but also within the same session as demonstrated by the significant effect of the block for all the parameters with the exception of correct letters, similarly for the interaction 'day' x 'block': task duration, correct letters, correct letters over task duration, extra action, see Table I.The effect of the group was present also in the typing task for all calculated parameters, see Table II.We also found a significant interaction group x day for the correct letters.Interestingly, when looking more in details at the distribution of performance with respect to age in the older adults group we found a linear relationship that linked the increase of age with a decrease in performance, see Fig. 5, both in the reaching and typing tasks (target hit: R 2 day1 = 0.56, p=0.046 and R 2 day6 = 0.68 p=0.021; correct letters/task time: R 2 day1 = 0.64, p=0.043 and R 2 day1 = 0.69, p=0.020).This relation is maintained also in the last day of training.
Results of the questionnaire (Fig. 6) highlight no significant differences between the two groups in all the answers to the 12 questions (p Q1 = 0.588, p Q2 = 0.835, p Q3 = 0.208, p Q4 = Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.

TABLE II RMANOVA EFFECTS OF THE BETWEEN GROUP FACTOR 'AGE', AND ITS INTERACTION WITH 'DAY' AND 'BLOCK' IN THE REACHING AND TYPING TASKS. MISSING VALUES REPRESENT NOT SIGNIFICANT INTERACTIONS
Fig. 6.Mean and standard deviation (error bars) of the answers to the 12 questions of the Post-Study System Usability Questionnaire (PSSUQ) for the younger (blue) and older (red) adults groups.Q1 to Q7 are about System Usefulness, Q8 to Q12 are questions about interface quality.

IV. DISCUSSION
This work contributes to the development of successful applications of MCI for intuitive use of assistive technologies, like a computer.The work focus on long term practice of people from different age groups.The developed MCI was proved effective through six 2-hour training sessions over two weeks for both younger and older adults groups that were trained to selectively use upper limb muscles to control the movements of a cursor on a computer screen completing a series of tasks.Key findings were the significant improvements in cursor control and therefore muscle coordination in both groups independently from age; all participants showed increased speed and accuracy after training although people in different age ranges exhibited various learning dynamics.Lastly, the questionnaire results regarding system usability and quality revealed a general consensus on the ease of use and intuitiveness of the interface among both younger and older adults.
More in details, with training all participants independently from age were able to improve their performance meaning that the quality of cursor control got better also in the group of older adults that might be less familiar in using technology and computers.Younger adults group achieve better results, acquiring since day 1 almost the totality of the presented targets in the reaching task and typing the entire sentence with no typing errors, decreasing the typing time each session.Older adults, despite the significant improvements in both evaluated tasks, in the last day were still missing few targets in the reaching task and made some errors needing extra corrective actions in the typing task.These differences in motor learning extracted with the fixed time design adopted in our study were not influenced by the fact that younger adults got more practice trials in the same time period compared to older adults.Indeed, when comparing the performance of younger and older adults for the same fixed number of presented targets, we obtain similar results (Fig. S1 in Supplementary material).The difference in performance between younger and older subject was expected based on a broad evidence of age-related changes in the motor and in particular muscular system.Studies on upper and lower limbs have shown changes to the peripheral nervous system that include a progressive loss of motor neurons [33], resulting in sarcopenia and a decline in strength.Additionally, with age, there is a tendency for limb musculature to exhibit slower contractile properties [29], [34].All of these physiological changes contribute to the widely reported reduction in motor coordination and fine control of forces and speed with age [26], [30], [35], [36].These factors influence certainly the outcomes of the training with our MCI being operated with multiple and coordinated muscle contractions, cursor movements were not only along a single direction, but it is encouraging that older adults group still reported improvements in cursor control.
Indeed, an interesting result is related to how age affected learning.Although learning was observable in both groups, the interactions between age and training performance highlighted that younger and older subjects had a significantly different learning speed.Indeed, the younger adults group started already from a better performance level and had a learning process that lasted for shorter period mostly occurring within the first day while the older adults group had a learning process that occurred until the last day.This interaction between age and training performance is highlighted also by the significant relationship between performance metrics and age within the older adults group (Fig. 5).Indeed the strong relation occurring in this group between age and motor performance at day 1 is becoming stronger, with a steeper regression line at day 6 meaning that the oldest subjects of the older adults group improved less.This is in accordance with what is reported in the literature about motor performance difficulties of older adults due to dysfunction of the central and peripheral nervous systems as well as the neuromuscular system and cognitive Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
abilities [26], [35], [36], [37], [38].Older adults are still able to learn gross motor skills but what is impacted with age is the abilities for finer motor tasks involving also bilateral coordination [28], [39].For instance a study on a complex task requiring coordination as the juggling task thought to youths, younger and older adults [39], in accordance with our findings, reported that older adults (60-79 years) performed significantly lower than 15-to 29-year-old youths and younger adults as well as they started from a poorer performance.Nevertheless the linear relationship between motor performance and age when people are older than 60 years old, to the best of our knowledge has been poorly investigated in the contest of a human-machine interface and it is worth investigating expanding the population of the older adults group recruited in this study.Despite the limited number of subjects used for the linear regression, that might lead to an overinterpretation of the correlation between performance and age, our preliminary results are encouraging and in accordance with Lee and Ranganathan [40] where they studied the early phase of learning with a body-machine interface across a wide age range and they found a rapid improvement until young adulthood and a decline con motor performance into the 50s-60s.They also demonstrated that performance were associated with limited exploration and the use of inefficient coordination patterns.
Finally, also the results of the questionnaire were positive.The fact that there was no group effect for the twelve answers is encouraging, meaning that the perception of the usefulness and quality of the developed MCI was not affected by the lower performance in our MCI use and by the different relationship with technology that older adults might have [7].The intuitiveness of the mapping, with a simple map transforming the activation of the four muscles into the four possible cursor movements helped operating the MCI and obtaining the achieved outcomes proving its good potential to be used for controlling also other external devices.Indeed, some devices such as keyboards, joysticks, or touch screen devices are not usable in situations where limbs cannot be used.People with motor disabilities, such as spinal cord injuries and neurological deficits following stroke or severe cases of multiple sclerosis, cannot manipulate a device directly, but with the help of assistive technologies, they can use technological tools to access information, social networks, work activities, or to control for example a powered wheelchair or another robotic device, thereby improving their autonomy and quality of life [10], [41], [42], [43], [44].Most of these studies are using residual kinematics while many researchers that investigated the use of EMG to control computer or wheelchair focused on facial EMG for controlling a cursor and many were using pattern recognition algorithm [45], [46].A drawback of such approaches is that as a pattern-based control scheme is used, only one pattern is accepted at a time and as a result, the cursor can only move in one of four directions while in this MCI the subjects could control the cursor along both directions at the same time.
Our proposed MCI offers potential advantages as it could be useful in the framework not only of assistive technologies but also of healthy ageing and neuromotor rehabilitation.Indeed, with aging populations worldwide, novel rehabilitation approaches are urgently needed to support health, function, and well-being in older adults and myoelectric interfaces provide a promising path forward.The choice of specific muscle can be customizable and adapted to the specific needs of each user.Moreover, some muscles could also be recorded and used not for control but for providing biofeedback when some negative compensatory strategies are ongoing, this would be very useful when working with people with neurological impairment following for example stroke or multiple sclerosis [20], [23], [47].Overall, the study provides valuable insights into the potential benefits of MCIs as an assistive technology although further research is needed to prove a significantly improvement of the quality of life for older adults.
The results of this study have implications for the field of healthy aging research and rehabilitation.As the world's population continues to age, finding effective interventions to promote healthy aging becomes increasingly important.MCIs represent a promising tool for achieving this goal.However, the small sample size of both groups tested in this work is a limitation that impact the generalizability of the findings.Moreover, additional metrics could be included to further analyze and characterize group difference and unveil whether the difference between older and younger adults is a matter of overall speed differences between the age groups or if there are other specific deficits in coordination.Additionally, more research is needed to determine the optimal frequency and duration of MCI training, and how to tailor MCI interventions to meet the specific needs and preferences of older adults.All the recruited participants are neurologically intact subjects, to understand the full potential of the interface people with neurological impairment should and will be tested in future studies.Further future development will also include the introduction of more captivating and challenging training games.

Fig. 1 .
Fig. 1.Experimental set-up of the muscle computer interface (MCI).The activity of right and left biceps (L BIC, R BIC) and right and left trapezius (R TRAP, L TRAP) were recorded with wireless sensors (green) and mapped into the x and y coordinates of the cursor.

Fig. 3 .
Fig. 3. Reaching metrics, mean and standard deviation (error bars), computed for the groups of younger (in blue) and older (in red) adults along the 6 days of training.In the light grey area are the performance computed from the first test block of the session while in the dark grey are the performance relative to the second and last test block.The vertical thick black line between day 3 and day 4 indicates the longer pause between training days due to the weekend.(a): Target hits and number of total targets (dashed line).(b): Number of targets missed.(c): Normalized target capture time (NTCT).

Fig. 4 .
Fig. 4. Typing metrics, mean and standard deviation (error bars), computed for the groups of younger (in blue) and older (in red) adults along the 6 days of training.In the light grey area are the performance computed from the first test block of the session while in the dark grey are the performance relative to the second and last test block.The vertical thick black line between day 3 and day 4 indicates the longer pause between training days due to the weekend.(a): Correct letters typed over the task duration.(b): Task duration.(c): Number of correct letters typed in the available time.(d): number of extra action corresponding to typing mistakes and corrections.

Fig. 5 .
Fig. 5. Relation between age and primary outcomes of the reaching (left panel) and of the typing (right panel) tasks.The blue and green dots refer to the performance of each subject of the younger (black outline) and of the older (no outline) adults group at day 1 and day 6 respectively, while the lines are the regression lines for the two days.The colored area indicates the mean±standard error of the performance of the younger adults.

TABLE I RMANOVA
EFFECTS OF 'DAY', 'BLOCK' AND THEIR INTERACTION IN THE REACHING AND TYPING TASKS