Age-related changes in movement control in virtual reality: a Fitts’ task protocol

Preserving motor and cognitive function in older adulthood is a central challenge for healthy aging and rehabilitation. Immersive virtual reality (VR) systems have emerged as promising tools to address this challenge, with many applications requiring users to perform targeted reaching movements toward virtual objects. Previous research suggests that aiming movements performed in immersive VR differ substantially from those executed in real-world settings, likely due to alterations in the multisensory information available for movement control. This is particularly relevant for older adults who experience changes in sensorimotor integration and slowing in information processing, which may make them more sensitive to altered sensory conditions. Despite the growing use of reaching tasks in VR, movement control during speed- and accuracy-constrained aiming in immersive contexts has not been systematically examined in older adults. Moreover, while augmented feedback is typically provided in VR reaching tasks, its implementation varies widely across the available studies. This limits our understanding of how different task and feedback conditions influence performance and the underlying sensorimotor control mechanisms in VR. To address this gap, we propose to examine the ability of healthy older adults (65–75 years) to perform Fitts’ task–a well-established, accuracy-constrained reaching paradigm–in immersive VR, in comparison with younger adults (18–28 years). To do so, task difficulty will be manipulated through target size to systematically impose varying accuracy constraints, thereby modulating cognitive control demands. Meanwhile, biomechanical constraints and VR-inherent perceptual factors, such as visual distortions and altered depth cues, will be held constant. Further, augmented feedback conditions will be varied to determine the best way to support accurate motor execution. The proposed protocol aims to provide a methodological benchmark for systematically evaluating motor performance in immersive VR and to support comparisons across populations and task conditions. The findings furthermore can inform the design of VR-based training and rehabilitation solutions for upper limb movements that aim to engage motor-cognitive processes in older adults.

Aging is associated with declines in both cognitive and motor functioning. Although the cognitive and motor domains were historically studied separately, cognitive processes and motor control processes are now widely recognized as functionally interdependent (e.g., Corbo, 2024; Heuninckx et al., 2005; 2008; Serrien et al., 2007; Zapparoli et al., 2022). In this view, increases in motor task demands–in both programming and online control of execution–result in greater recruitment of cognitive resources in both young adults (YA) and older adults (OA) (Corbo, 2024; Zapparoli et al., 2022). Notably, aging results in a stronger association between both domains (motor-cognitive interactions), suggesting that OA increasingly rely on cognitive resources for successful motor execution (Heuninckx et al., 2005; 2008). Evidence supporting this hypothesis has been reported across a range of paradigms, including walking (Langeard et al., 2019) and bimanual coordination (Temprado et al., 2020). Moreover, brain imaging studies showed alterations of brain activation during aging, such as the Posterior-Anterior Shift in Aging (Davis et al., 2008; Zapparoli et al., 2022), neural dedifferentiation, and/or compensatory overactivation of less specialized circuits (Heuninckx et al., 2005; 2008).

Since immersive VR systems are becoming increasingly integrated into motor rehabilitation and gaming (e.g., Gu et al., 2024; Kirthika, 2025; Preikstas and Schofield, 2025), the question arises however of whether and how information processing demands and motor-cognitive interactions are impacted when motor tasks are performed in virtual reality (VR) environments. In the present work, which addresses this issue, immersive VR refers to computer-generated 3D environments that users experience through a head-mounted display (HMD). The virtual representation is spatially realistic and allows the users to perceive and interact with virtual objects. All other virtual environments, such as desktop displays with stereoscopic 3D glasses are considered low-immersive (for similar definitions, see, e.g., Kaplan-Rakowski and Gruber, 2019). VR is currently considered a promising tool for motor assessment, training and rehabilitation since it enables flexible task design, including: (i) manipulation of spatial parameters and informational context (e.g., augmented feedback [FB], structured environments), (ii) gamification of exercises, and (iii) simulation of realistic and ecologically relevant motor tasks (Calderone et al., 2025; Temprado, 2021). In this context, VR can serve not only to assess movement control mechanisms but also as a tool for delivering motor tasks in clinical settings. However, before VR-based assessments or interventions can be optimized and applied to rehabilitation programs, it is necessary to understand how the VR environment itself influences motor behavior across diverse populations. To date, immersive VR is often used primarily to increase patient engagement and adherence to therapeutic exercises through gamified interactions, in healthy populations (Mader et al., 2016) but also in other populations (e.g., stroke patients; Zuki, 2021).

Notably, however, evidence suggests that movements performed in VR can differ from those executed in real-world settings. Previous studies using non-immersive VR (specifically, desktop displays with stereoscopic 3D glasses) have reported longer movement times, more sub-movements, and reduced spatial accuracy during reaching tasks, particularly in the anterior-posterior direction (Liu, 2013; Barrera Machuca and Stuerzlinger, 2019). However, the extent to which these findings generalize to immersive VR remains unclear. Recent work using immersive VR suggests that some of these effects may extend to HMD-based contexts, especially regarding reduced spatial accuracy compared with non-VR conditions, again particularly along the anterior-posterior axis (McAnally et al., 2023; McAnally and Wallis, 2024). Within immersive VR, users are presented with a modified sensory environment including changes in visual, proprioceptive, and vestibular information (Bhowmik, 2024). Although HMD-based immersive VR has advanced substantially to support depth perception and spatial interaction (Gerig et al., 2018), these differences in sensory information may still disrupt multisensory integration, particularly between visual and proprioceptive signals (Pratviel et al., 2021), which is essential for accurate movement control. These alterations may be more detrimental for OA (Kim et al., 2024), as aging is associated with greater reliance on multisensory integration and, specifically, on visual input for movement guidance (de Dieuleveult et al., 2017; Seidler et al., 2010). However, the effects of immersive VR on movement control remain insufficiently characterized, particularly in OA, with only a few studies beginning to address this question (e.g., Gu et al., 2024; Kim et al., 2024; Wu et al., 2024). Thus, movement control in VR may impose greater information processing demands than in real-world environments, particularly in OA and as task constraints become more demanding. This hypothesis is in line with the results of a recent study showing that reaching tasks performed in VR are associated with high cortical activation in perceptual brain areas, particularly in OA (Gu et al., 2024), potentially reflecting greater reliance on sensory feedback during movement execution. Together, these findings suggest that movement control may be particularly sensitive to VR contexts, highlighting both the potential and the methodological challenges of using immersive VR to investigate and stimulate motor-cognitive interactions during movement control, especially in aging. This underscores the need for controlled experimental paradigms enabling comparisons across different task constraints, populations, and VR configurations. The present study is a step in this direction.

One potential approach to mitigate the effects of an altered informational context in immersive VR is the use of augmented FB, specifically knowledge of results. Knowledge of results provides information about the outcome of a movement after its completion (rather than information about the movement pattern itself; i.e., knowledge of performance) and can support performance by allowing participants to detect errors and adjust subsequent movements on a trial-to-trial basis (Salmoni et al., 1984). In the present study, augmented FB in the form of knowledge of results will be provided over a relatively small number of trials, as usual in motor performance experiments. The role of knowledge of results in movement control has been widely investigated in real-world contexts (Salmoni et al., 1984; Schmidt et al., 2019), including several studies on OA (Liu et al., 2013; Swinnen et al., 1998; van Dijk et al., 2007). Overall, these studies suggest that knowledge of results can support movement performance in both YA and OA, with inconsistent findings on age-related differences. In VR, its role remains largely unexplored with most studies not including control conditions without FB (e.g., Batmaz, 2019; 2020; Batmaz and Stuerzlinger, 2022; Kourtesis et al., 2022; Tanacar, 2023). Nevertheless, emerging evidence suggests that terminal FB can support accurate task performance in immersive VR. In this sense, McAnally and Wallis (2024) reported that terminal auditory error FB significantly reduced aiming errors compared to a no FB condition in VR-based rapid target-directed aiming tasks.

Taken together, these findings raise the questions of whether and how: i) the control of rapid target-directed movements is affected when performed in immersive VR environments, particularly in OA and ii) terminal augmented FB optimizes performance.

To address these issues, it is necessary to examine motor performance under systematically varied task demands. One approach to manipulating such demands is through Fitts’ task, which provides a well-established paradigm for studying how objectively quantified task difficulty influences movement control (Fitts, 1954; Fitts and Peterson, 1964). The discrete version of Fitts’ task requires participants to perform rapid target-directed movements from a fixed starting point to a target as quickly and accurately as possible (Fitts, 1954; MacKenzie and Buxton, 1992). Task difficulty can be systematically manipulated by varying target distance and/or target width, such that increasing target distance or decreasing target width increases task difficulty. Difficulty is quantified using the index of difficulty (ID), which combines target distance and width into a single metric defined as log 2 2 * D W. According to Fitts’ Law, movement time (MT) increases linearly with IDs. This relationship is summarized by the MT-ID efficiency function, whose slope primarily reflects the efficiency with which participants can process task-relevant information as speed-accuracy demands increase (information-processing efficiency). Steeper slopes indicate larger increases in MT as task difficulty rises and are interpreted as lower information processing efficiency, whereas shallower slopes indicate more efficient processing. Importantly, MT-ID functions can also be influenced by biomechanical and spatial factors, such as movement direction along the three spatial axes (up/down, left/right, forward/backward; Cha and Myung, 2013; Murata and Iwase, 2001), the number of degrees of freedom involved (such as shoulder rotation and elbow elevation; Morasso, 1983), and, presumably, the aforementioned VR-specific sensory alterations, including visual distortions. By keeping movement distance, direction, and VR setup constant across trials, the present study aims to minimize these additional influences in the assessments of MT-ID functions as a measure of task-related information processing efficiency in both YA and OA.

Based on kinematic analysis of the different phases of movement execution–acceleration and deceleration, separated by peak velocity–a previous study carried out in our research group suggested that information processing demand, a proxy of attentional processes involved in movement control, is primarily reflected in the deceleration phase, which relies on online adjustments of the movement trajectory to achieve terminal accuracy (Temprado et al., 2013). Consistent with this interpretation, deceleration time has been shown to scale with target size, with smaller targets requiring greater precision and longer deceleration phases (MacKenzie et al., 1987; Temprado et al., 2013). Sleimen-Malkoun et al. (2013) observed similar EF slopes in the Hick-Hyman task, a cognitive choice reaction time paradigm, as in the deceleration phase in Fitts’ task in OA, indicating comparable age-related slowing of information processing across primarily motor and cognitive tasks. This pattern suggests that the deceleration phase of Fitts’ task may reflect shared processing capacities that limit performance across the motor and cognitive domains. Fitts’ task therefore provides a robust paradigm for assessing movement control under systematically manipulated task demands and for making inferences about motor-cognitive interactions, including in VR.

Research gaps and objectives of the present experiment

The present experiment is designed to address several remaining questions related to the effects of aging and provided FB on reaching control in immersive environments. First, while VR-based reaching movements often show altered kinematics, it remains unclear how information processing demands can be systematically elicited through accuracy constraints while controlling biomechanical and VR-specific factors. Second, age-related differences in movement control during VR-based Fitts’ task performance remain largely unexplored. Finally, the role of augmented terminal FB in supporting error detection and movement execution during accuracy-constrained reaching tasks in VR is not fully understood.

To address these questions, the present study will investigate movement control in immersive VR by manipulating task difficulty and FB, and by examining age-related differences using a discrete Fitts’ task. Target size will be manipulated to yield different IDs, thereby varying the amount of information that must be processed to achieve movement accuracy. Specifically, we will compare movement kinematics and error rates across different conditions, including a condition in which speed is explicitly prioritized. We hypothesize that MTs will be shorter and error rates higher in this latter condition than in the other conditions involving standard Fitts’ task instructions requiring a speed-accuracy tradeoff. Second, we will assess whether Fitts’ Law holds in VR when only inherent visual FB is available, as in real-world conditions. In this context, we aim to determine whether participants can successfully perform Fitts’ task without augmented FB. We hypothesize that, in the absence of augmented FB, participants will rely only on inherent sensory information to evaluate movement outcome. Because inherent sensory information is altered in immersive VR and may therefore be insufficient to reliably detect endpoint deviations, particularly under high accuracy demands (i.e., higher IDs, smaller targets), participants may be less able to adjust their movements on a trial-to-trial basis. Consequently, we expect them to execute quicker movements and miss the target more often than in conditions with augmented FB. Third, we will examine whether augmented visual FB promotes behavior consistent with Fitts’ Law. To this end, we will compare two FB types that differ in informational value (i.e., global and specific FB; see Methods). We hypothesize that augmented FB will reduce MTs and error rates relative to the inherent FB condition, yielding a typical linear increase in MT with ID. Moreover, FB with higher informational value is expected to result in greater benefits. Finally, we will investigate age-related differences in task performance by comparing YA and healthy OA. Our hypothesis is that OA will exhibit longer MTs, higher error rates, steeper EF slopes, and disproportionately longer deceleration phases. Further, we assume that both YA and OA may benefit from augmented FB–particularly FB providing more specific information about the error type–due to its potential to compensate for changes in sensory information in immersive VR which might, in turn, affect movement control. Potentially, we may find increased benefits from augmented FB in OA relative to YA as it may compensate for age-related performance deficits, especially those due to changes of multisensory integration–but this remains speculative. Primary outcome will be interaction effects of age groups and experimental conditions on the slopes of the MT-ID, AT-ID, and DT-ID regression functions. Secondary outcomes include the effects of FB condition and age group (and their interactions) on performance metrics: MT, AT, DT, and error rate.

For clarity, the abbreviations used throughout this article are listed in Table 1.

Participants

Thirty participants – 15 YA (18–28 years) and 15 OA (65–75 years) – will be recruited. The sample size was determined via a power analysis using G*Power for a mixed-design repeated-measures ANOVA (Faul et al., 2007), focusing on our primary outcome (Condition x Age Group interaction in slopes of MT-ID, AT-ID, and DT-ID efficiency functions). The analysis assumed an alpha level of 0.05, statistical power of 0.80, a medium effect size (f = 0.26), a correlation among repeated measures of 0.5, and a non-sphericity correction factor of 1.0. To our knowledge, no VR Fitts’ task studies have analyzed EF slopes to date. However, because throughput is approximately the inverse of EF slope, we based our effect size estimate on throughput studies, which provide the closest available empirical basis for estimating an expected effect. Specifically, Batmaz and Stuerzlinger (2021) reported a within-subject interaction of selection technique × auditory FB on throughput (partial η 2 = 0.22, f = 0.53). Because the published effect was obtained for a purely within-subject interaction on throughput, whereas the present study tests a mixed interaction that is expected to exhibit greater between-participant variability, we adopted a substantially more conservative effect size (f = 0.26) for the power analysis. Since this yielded an estimated required total sample of 22 participants, recruiting a total of 30 provides a margin for potential attrition or unusable data and aligns with sample sizes in similar VR studies.

OA will be recruited from an existing cohort of community-dwelling volunteers and supplemented via local advertisements and senior activity centers, while YA will be recruited from the university campus through posters, mailing lists, and word-of-mouth.

All participants must be right-handed, have full functional range of motion in the dominant arm, normal or corrected-to-normal vision, no color blindness, have no history of neurological or psychiatric disorders, and have normal cognitive functioning; this will be self-declarative in YA, whereas OA will be screened for normal cognitive function using MoCA questionnaire (cut-off value ≥26) due to their age and related risk of cognitive decline. Self-reported exclusion criteria include frequent video game use (>5 h/week) – independent from type or platform of gaming, to reduce potential confounding effects of prior gaming experience on visuomotor performance (e.g., Gugura et al., 2023) –, uncorrected sensory or motor impairments, neurodegenerative disease, or medications affecting cognitive or motor function. Additional exclusion criteria assessed during the experiment include severe motion sickness or intolerance to VR, as well as inability to properly adjust or comfortably wear the VR headset. Furthermore, all participants will be screened for visually induced motion sickness susceptibility using a dedicated questionnaire (VIMSSQ with cut-off ≥12; Golding et al., 2021; Golding and Keshavarz, 2025) and participants’ self-reported functional limitations will be obtained.

Recruitment will take place over approximately 2 months, with informed consent obtained from all participants in French or English. All participants will provide written informed consent prior to participating in the study. The protocol was approved by the French ethic committee CERSTAPS (IRB00012476-2025–25-11–451) and is therefore in accordance with the ethical standards laid down in the Declaration of Helsinki.

VR set-up

The VR setup (see Figure 1) was developed through a close collaboration between human movement scientists from the Institute des Sciences du Mouvement and VR developers from the Centre de Réalité Virtuelle de la Méditerranée, both at Aix-Marseille University. Based on iterative testing with pilot participants, we optimized the UI and UX design of the setup to improve usability, particularly in OA. Thus, the interface became increasingly intuitive, allowing both young and older participants with little or no VR experience to easily interact with it, after few familiarization trials. Notably, the environment is not gamified to maintain a neutral, laboratory-like setting consistent with previous Fitts’ task studies in both VR and real-world contexts.

Commercially available equipment will be used: a PICO 4 Ultra Enterprise HMD, PICO Ultra controllers, and a NVIDIA RTX 5070 GPU computer system. To our knowledge, no studies are yet available examining the tracking accuracy and reliability of the PICO 4 Ultra Enterprise. However, recent work using the Meta Quest 2, Quest 3, and Quest Pro, which are based on comparable inside-out optical controller tracking systems, demonstrated good agreement with gold-standard optical motion capture systems (Vicon and Qualisys) for upper-limb kinematic assessment (Costa et al., 2026; Fabrizio et al., 2026). Notably, tracking performance was highest when the controllers remained within the HMD’s field of view. As all reaching movements in the present protocol are performed within the field of view, we expect high controller visibility throughout the experiment. In addition, several recent studies have successfully employed the PICO 4 Ultra Enterprise or its predecessor in experimental research, supporting its feasibility and practical applicability for the proposed context (van den Hoek Ostende et al., 2025; Wu et al., 2025). The scenario was developed in Unity 6000.0.37f software using a Windows 11 (64-bit) PC with Intel® Core™ i5-13500H processor, 16 GB RAM.

Task and conditions

Throughout the experiment, participants will be seated comfortably in a height-adjustable seat, with their back leaning against a backrest, in front of a desk. Chair height will be adjusted so that participants can maintain an ergonomically comfortable posture, comparable to that used during writing, allowing the hand to rest on the desk between trials. Participants’ feet will be positioned flat on the floor, with a footrest provided when necessary. In the virtual environment, the virtual desk and apparatus will be aligned with the real setup so that the starting point–a virtual 3D sphere–appears directly in front of the participant. Movements will be performed while grasping a handheld controller in their right, dominant hand which is tracked by the HMD’s built-in system. The HMD will be worn in all conditions. Participants will perform reaching movements with the controller from a fixed starting location (a sphere of 3 cm diameter) to a target positioned along the body’s sagittal midline. The controller will be rendered in the virtual environment, with a small white point displayed just in front of the controller indicating its tip and serving as the reference point throughout the experiment. This tip must enter the starting and target spheres.

Before each trial, participants will rest their right hand on the desk while pointing the tip of the controller at the start sphere. The center of the targets will be placed 20 cm forward and 15 cm upward (vertical) from the center of the start position, yielding a resultant distance of 25 cm (according to Pythagoras’ theorem a 2 = b 2 + c 2), whatever the size of the target. Prior to the experiment, the setup will be calibrated in a way so the movement will at most correspond to about 80% of maximal forward reach. The start and the target spheres will be represented as full grey spheres with a grid-like surface, making their 3D structure visible. Target diameters will correspond to the different target width values (see below) while the start sphere will have a diameter of 3.00 cm.

To fix the IDs, we used the initial calculation formula proposed by Fitts (1954), which has been shown to remain valid in VR (Preikstas and Schofield, 2025). Accordingly, ID will be manipulated via target diameters (i.e., equivalent of target width in 2D ID calculation) of 12.50 cm, 6.25 cm, 3.13 cm, and 1.56 cm, yielding IDs of 2, 3, 4, and 5 bits, respectively (see Figure 2).

Each trial will begin with the controller tip resting in the start sphere. Participants are then informed that they can initiate the movement when ready and instructed to reach the center of the target as quickly and accurately as possible, remaining inside the target sphere for 0.5 s to validate the end of the trial.

The system will record several types of errors, including directional deviations (to the right or left of the target as well as above or below the target) and amplitude errors in which the movement was too short or too long. A trial will also be classified as an error if the participant entered the target and exited again before the end of the validation period, i.e., not remaining within the sphere for 0.5 s.

Four conditions (presented below) will be compared: inherent FB (iFB), global FB (gFB), specific FB (spFB), and a condition with instructions to prioritize speed (iFB-speed). The iFB-speed condition will serve as a control condition with respect to the classical Fitts’ task instructions (i.e., the speed-accuracy trade-off), whereas the iFB condition will serve as a control condition for comparisons between the different types of augmented terminal FB (see Figure 3).

In the iFB-speed condition, the immersive VR environment will be presented without augmented FB (i.e., inherent FB only), and participants will be instructed to reach the target as quickly as possible without prioritizing accuracy. Beyond examining movement control under speed-focused instructions, this condition will serve to verify that participants comply with the standard speed-accuracy trade-off instructions in the other conditions, in which participants will be required to move as quickly as possible while also ensuring accurate target acquisition. In the iFB condition, again only inherent FB will be available. In the gFB condition, augmented visual FB will indicate target entry and trial outcome: the target turns blue when entered, green after participants remain inside for 0.5 s, and red in case of an error (either directly if not entered or after briefly turning blue if entered and then exited). In the spFB condition, additional augmented FB will provide information about trial outcome via a written message displayed after each trial, specifying the type of error when applicable or indicating successful target acquisition (‘success’). As endpoint errors often involved multiple directional components rather than a single dimension, six error regions were defined around the target, with FB corresponding to the predominant directional deviation. The messages will display: ‘too long’ for overshoot, ‘too short’ for undershoot, ‘error to the right/left’ for lateral deviations, and ‘too high/low’ for vertical deviations. During the design phase, quantitative error FB (e.g., “2 cm too long”) was considered; however, pilot testing suggested that this level of detail for errors combining multiple directional components (e.g., “2 cm too long and 1 cm too far to the left”) became cognitively demanding and less intuitive for participants, potentially reducing rather than improving the usefulness of the FB. Therefore, qualitative FB was retained.

Experimental procedure

For an overview of the procedure, see Table 2. Before the experimental session, participants will complete a familiarization phase to become acquainted with the VR environment and task instructions. They will first perform practice trials at an ID of 2.5 (target size: 8.84 cm; movement distance: 25), followed by practice trials at an ID of 4.5 (target size: 2.21 cm; movement distance: 25 cm). The IDs used will differ from those included in the experimental session to minimize potential learning effects. The practice trials will be conducted with specific FB, as it provides the greatest amount of task-relevant information to support understanding of the task and the FB. Participants will be encouraged to ask questions throughout the familiarization phase. The number of practice trials will be individually determined, with the familiarization ending once participants indicate that they understand the task instructions and the provided FB.

Throughout the experimental session, the order of the four conditions will be counterbalanced to avoid order effects and within each condition, the order of the four IDs (2, 3, 4, and 5) will be randomized across blocks of trials (see Figure 4).

In total, participants will perform 15 trials per ID per condition, yielding a total of 4 × 4 × 15 = 240 trials. Following the familiarization phase and between each experimental condition, participants will be asked whether they wish to take a break. They will also be encouraged to take additional breaks whenever needed and will furthermore be instructed to rest their arm on the desk in front of them between trials to avoid eventual fatigue throughout the experiment. Given these fatigue mitigation strategies and the relatively low physical demands of the task, we do not anticipate substantial fatigue effects.

Data recording

The 3D positional signal (x-, y-, and z-coordinated) of the handheld controller will be recorded using the HMD’s built-in tracking system at a constant sampling frequency of 100 Hz and saved for further processing. In addition, trial outcome will be recorded for each trial, indicating whether the movement endpoint was successfully acquired within the target area or resulted in an error. The acquired positional data will be pre-processed before kinematic analyses. In order to determine the appropriate filtering procedure, spectral analysis of the raw trajectories will be performed to determine the frequency characteristics of the reaching movements and identify potential contributions of tracking noise. Moreover, raw and filtered signals will be visually inspected to ensure that filtering reduces noise while preserving relevant movement characteristics. Based on previous kinematic analyses of VR-based reaching movements, a low-order (second-to fourth-order) dual-pass (no phase lag) Butterworth filter with a cut-off frequency in the range of 1–10 Hz is expected to be appropriate (Clark et al., 2024; Culp et al., 2024; de Nobile et al., 2025; Fabrizio et al., 2026; Fu et al., 2012). Velocity profiles will then be computed from the filtered positional data.

Threshold values for movement onset and offset were initially based on prior literature (Teasdale et al., 1993) and subsequently refined through pilot testing to account for the noise characteristics of VR-based tracking systems and small involuntary hand movements associated with physiological tremor (Andrade, 2013). Based on iterative analyses exploring different velocity thresholds, movement onset and offset were defined as the first frame in which velocity exceeds 12.5 cm/s and the first frame in which it remains below this threshold for at least 0.5 s, respectively (see Temprado et al., 2013, for a similar procedure in real-world conditions). Combined with the analysis of the spatial position, this procedure helps ensure that minor velocity fluctuations arising from VR tracking noise and physiological tremor are not misinterpreted as unfinished movement.

Three main variables will be calculated for each trial: (1) MT, defined as the time elapsed between movement onset and offset; (2) acceleration time (AT), defined as the duration from movement onset to peak velocity; and (3) deceleration time (DT), defined as the duration from peak velocity to movement offset. In addition, for each participant and ID, the effective target width will be calculated from the standard deviation of movement endpoints (for a detailed procedural description, see MacKenzie, 1992; MacKenzie, 2018; Sleimen-Malkoun et al., 2012), reflecting the actual precision achieved by participants during the task. To verify whether the prescribed IDs were respected, the spatial distribution of movement endpoints will be compared to the target-defined constraints (i.e., target distance and width), following Sleimen-Malkoun et al. (2012). Additional performance metrics will be computed, namely, error rate, defined as the proportion of missed targets relative to the total number of trials, and linear regressions of MT, AT, and DT against ID.

Finally, participants will rate their perceived cognitive effort and perceived exertion using simplified versions of the Paas scale (Paas et al., 1994; 2008) and the Borg CR10 scale (Borg, 1982; 2008), respectively. Both ratings will be collected on a 0–10 Likert-type scale, with five representing a moderate level, for consistency and ease of use. Cognitive effort will be assessed after each target size within each experimental condition to capture variations across both task difficulty and condition. Physical exertion, focused on the shoulder and arm, will be assessed after the familiarization block and at the end of each experimental condition to track changes over the duration of the experiment.

Data analysis

Analyses will be conducted using repeated-measures and mixed-design ANOVAs, reporting main and interaction effects, unless otherwise specified. Normality and homogeneity of variances will be assessed using the Shapiro-Wilk and Mauchly’s tests, respectively. If the assumption of sphericity is violated, Greenhouse-Geisser corrections will be applied. If normality assumptions are not met, Friedman tests will be used as a non-parametric alternative. Post hoc comparisons will be performed where appropriate using Bonferroni corrections, and partial eta squared (η 2 p) will be reported for all effects. Across all analyses, statistical significance will be set at p <.05. Age group (2 levels: YA and OA) will be treated as a between-subject factor, and all experimental factors as within-subject factors unless otherwise specified. Only correct trials will be included in analyses, except for error rates.

Performance variables will be computed as follows: For MT, AT, and DT, mean values will be computed for each participant, prescribed ID, and condition based on all correct trials per ID. Error rates will be calculated as the proportion of incorrect responses across all trials per ID for each participant and condition. Each variable will then be analysed using separate ANOVAs with condition (4 levels: iFB-speed, iFB, gFB, and spFB) and age group as factors to assess performance across conditions and a potential aging effect. Second, to verify whether Fitts’ Law holds across experimental conditions, linear regressions will be performed separately for each participant and condition for each temporal variable (MT, AT, and DT). For each participant x condition x temporal variable combination, mean values per prescribed ID will be computed, yielding four data points per regression. Slope, coefficient of determination (R 2), and statistical significance will be extracted for each participant and condition. Then, regression slopes for MT-ID, AT-ID, and DT-ID will be analyzed separately using ANOVAs with condition and age group as factors. To examine potential trial-to-trial familiarization or adaptation effects, all performance analyses will be repeated after excluding the first two trials of each experimental condition. In addition, potential changes in performance over the course of each block of trials within each condition will be explored post hoc by comparing performance during the first and last four trials of each condition.

Physical exertion (focused on the shoulder and arm) will be rated after the familiarization block and after each experimental condition. To evaluate changes in perceived exertion over the course of the experiment, Borg scores will be analyzed using a 2-way mixed-design ANOVA with timepoint (5 levels: familiarization and four conditions in order of presentation) and age group as factors, allowing assessment of physical fatigue across the experimental session. Participants will further rate their cognitive effort after each ID within each condition. To assess the effect of condition, mean cognitive effort scores will be computed for each participant and condition by averaging across IDs. To assess the effect of task difficulty, mean scores will be computed for each participant and ID by averaging across conditions. Condition effects and ID effects will then be analyzed using separate ANOVAs with age group as a factor in both cases.

For an overview of every dependent and independent variable, see Table 3.

We expect the iFB-speed condition to yield the shortest MTs, the highest error rates, and the shortest deceleration phases. In this condition, regression slopes of MT-ID, AT-ID, and DT-ID are expected to be flat or show random patterns. In the iFB condition, in which, again no augmented FB is available but this time, both speed and accuracy demands are present, MTs and specifically DTs are expected to increase, and error rates to decrease relative to the iFB-speed condition. However, MT-ID, AT-ID, and DT-ID slopes are expected to remain relatively flat. In contrast, the gFB and spFB conditions are expected to promote more accurate performance with lower error rates compared to the iFB condition, while MTs–and disproportionally DTs–are expected to further increase. Potentially, this effect may be more pronounced in the spFB condition. In both conditions with augmented FB, MT-ID, AT-ID, and DT-ID functions are expected to show a clear linear increase with ID, consistent with Fitts’ Law.

Age-related differences are expected across conditions, with OA exhibiting longer MTs, disproportionately longer DTs, and, potentially, steeper MT-ID, AT-ID, and DT-ID slopes compared to YA in the conditions with augmented FB. We speculate that the effect of augmented FB (gFB and particularly spFB) shifting behavior towards longer MTs and DTs and lower error rates might be more pronounced in OA.

Finally, perceived cognitive effort (Paas scale) is expected to increase with task difficulty (i.e., higher IDs) and, potentially, over time. Perceived physical exertion is expected to increase over time, although remaining below moderate values (<7).

Given the increasing use of VR technology in motor training and rehabilitation, it is important to understand how well fundamental models of movement control, such as Fitts’ Law, apply in these immersive environments for interpreting motor performance and designing effective VR-based applications. This is particularly relevant for OA, who represent a primary target group for VR-based solutions for rehabilitation and training.

Beyond evaluating the applicability of Fitts’ Law, the proposed protocol provides a standardized framework for quantifying MT and, in particular, DT under speed-accuracy constraints. These measures may offer valuable insights as indicators of information processing demands and attentional load during task execution, enabling inferences about motor-cognitive interactions under different FB conditions. If results are consistent with our predictions, the present study will show that Fitts’ Law is valid for target-directed aiming in 3D immersive VR across both YA and OA. In addition, it will provide a means to assess age-related changes in information processing during unimanual reaching movements in immersive VR. Finally, the present protocol could provide a standardized framework for evaluating how different augmented FB modalities, task designs, and future technological progress in the development of training and rehabilitation solutions in immersive VR influence movement speed, accuracy, and information processing demands. Indeed, as advances in VR technologies continue to enhance the realism of the immersed experience (Bhowmik, 2024), the present protocol could help determine whether these developments result in improvements in motor performance. In rehabilitation contexts, our protocol could further support the tailored selection of task difficulty and FB conditions that correspond to the abilities and needs of specific user populations, including but not limited to OA.

Limitations

A limitation of the present protocol could be the absence of a dedicated baseline assessment of motor performance outside the VR environment, which would enable a direct comparison between real-world and VR-based reaching performance. Moreover, other potential covariates will not be obtained, namely, prior VR experience, affinity for technology, and educational level. Forthcoming studies should obtain these variables and incorporate task-based motor and cognitive baselines assessed outside VR. However, given the range of covariates assessed in the context of this study, this is not expected to compromise the study’s objectives, the validity of the findings or its contribution.