The integration of perception and control is crucial for enhancing the performance of robotic systems in complex and dynamic environments. Immersive virtual reality (VR) enables users to experience a sense of presence (Rose et al., 2018; Kuhne et al., 2023; Alazmi and Alemtairy, 2024; Pallavicini et al., 2019), making it an ideal tool for studying motor function and control, which are critical for designing effective actuators in robotics. VR environments offer practical advantages in a three-dimensional (3D) settings (Knight and Titov, 2009; Anoffo et al., 2018; Setti and Csapo, 2021) while few studies have been conducted in a non-immersive two-dimensional (2D) environment (Tam et al., 2005). In general, the quality of VR can change user behavior (Lee et al., 2020; Hoffman et al., 2006; Rosa et al., 2016). Creating accurate VR environments helps us understand human motor control and perception.

Reaching has been widely used to evaluate motor function (Roby-Brami et al., 2003; Cirstea et al., 2003; Igarashi et al., 2011; Kim et al., 2020a) because it is a fundamental task in daily life and can include other complex movements. Deployed in VR environments, reaching performance has been evaluated to investigate the effect of different viewing medium (Subramanian and Levin, 2011; Lin et al., 2019) or different displays (Camporesi and Kallmann, 2016). These previous studies successfully indicated that reaching performance could vary depending on visual representation. However, a task for 2D movement does not necessarily need to adopt complex 3D environments. Although planar reaching movements can be easily represented on a monitor using simple cursors (Kim et al., 2019; Brière and Proteau, 2011; Onagawa et al., 2022; de Brouwer and Spering, 2022; Ikegami et al., 2022), the experience provided by VR involving 3D objects is fundamentally different.

The 3D space enables users to perceive more information than the 2D space, leading to different perceptions of visual complexity because of depth information (Li et al., 2020), greater emotional arousal caused by binocular disparity (Tian et al., 2021) and different grip performances that depend on the view angle (Keefe and Watt, 2017), than in 2D environments. Several factors improve the quality of the 3D space and are associated with the mode of 3D information processing by the brain. Although some of these factors contribute to a better appearance, they are not indispensable in enabling people to recognize a VR environment as a 3D space. Thus, as an indispensable factor, we focused on the stereoscopic effects of objects in this study, breaking the 3D environment down into several aspects that compose it. Some previous studies have reported that stereoscopic effects in VR environments are negligible in public speaking (Ling et al., 2012) and in acquiring knowledge about heart disease by watching 2D or stereoscopic heart models (Patel et al., 2021). Nonetheless, the evaluation indicators used were too abstract to investigate the sophisticated role of stereoscopic objects in general immersive VR environments, calling for an investigation based on behaviors in more general and basic tasks to determine whether stereoscopic properties should be considered for creating an immersive VR environment for motor function assessment.

Therefore, in this study, we investigated the effects of stereoscopic objects during a reaching task in a systematical way by observing behavioral performance, examining whether the stereoscopic effect of objects is a significant aspect enough to affect behavioral level. We hypothesized that the stereoscopic effect of reaching objects would influence the reaching performance. As the stereoscopic property is processed for depth perception in the brain (Backus et al., 2001; Guo et al., 2022), cognitive function should be related to the performance, linked to age as a critical factor for the function (Deary et al., 2009). Thus, young and older individuals were recruited separately for the experiment. We adopted 2D and 3D objects as the cursor and target in a reaching task requiring the same reaching trajectory. Moreover, we assessed distance to the target, which is associated with difficulty of a task (Fitts, 1954), as two conditions (long and short distances) for the experiment. We expect the results to determine rough effects of the factors.

First, we compared end-point errors using feedforward control between the young and older age groups. Figure 2 shows the mean end-point error while comparing the young and older age groups for each condition. For all tasks, the mean end-point error of the young group was less than that of the older group. However, the difference in error was significant (p < 0.001, effect size |r| = 0.59) only in the 3D long task, where the mean errors for the two groups were 0.009 and 0.032 [m], respectively. The errors in the 2D long task were 0.008 and 0.018 [m], respectively. Tasks that required a long distance were associated with a large difference between the young and older groups.

Figure 3 shows the smoothness of movement. In all tasks, the young group was significantly better in smoothness than the older group (p < 0.001). The Cohen’s d values for each task were 1.42, 1.57, 2.21, and 1.32, respectively. Consistent with end-point error, long tasks showed a significant difference. The values for smoothness in the young and older groups for the 2D objects were −1.507 and −1.743, respectively, and those for the 3D objects were −1.563 and −1.799, respectively. These data suggest that smoothness can be used for more accurate assessment of performance than end-point error.

Because we found that the smoothness parameter was appropriate for evaluating movement performance, we investigated whether stereoscopic objects would make a difference. Figure 4 shows the smoothness of movement in the young and older groups. Although the differences between 2D and 3D objects for the long tasks were larger than those for the short tasks, the difference was significant only for the long task by the young group (p < 0.05, Cohen’s d = 0.85). The mean smoothness in the long task was −1.506 (2D) and −1.563 (3D) for the young group and −1.743 (2D) and −1.799 (3D) for the older group. For both young and older groups, the difference in smoothness between 2D and 3D objects for the short task was almost negligible. No group showed a significant difference in the short task, suggesting the ceiling effect by distance.

In this study, we investigated whether stereoscopic objects affect reaching performance. Our results revealed that the older group had significantly worse predictive control than the young group in the 3D long task, suggesting that the positions of 3D objects were more difficult to predict than those of 2D objects. The young group showed a significant difference in smoothness between 2D and 3D objects, which was not observed in the older group. Given that older adults are reported to experience a decline in binocular vision and stereopsis, which are crucial for depth perception (Norman et al., 2009), the results for the older group may reflect a floor effect. Under the short-distance condition, neither group showed a significant difference, suggesting the ceiling effect by distance. We confirmed that the effect of stereoscopic objects was not drastic but it did hamper the reaching performance.

In a different situation, it was reported that an immersive stereoscopic picture impeded people from learning vocabulary and that learning with 2D images was more effective (Kaplan-Rakowski et al., 2022). The authors conjectured that the immersive nature of stereoscopic pictures was a distractor or drained some of the cognitive resources required for learning. Similarly, the perception of stereoscopic objects may necessitate additional information processing in the brain, which is not necessary when a 2D object is perceived. Considering this, prior experience with 2D or 3D devices could potentially influence participants’ performance. Although we did not collect information on participants’ prior exposure to these devices, it would be worthwhile for future studies to investigate whether such exposure affects performance. Several factors contribute to the perception of an object as stereoscopic. For example, shading is important for perceiving the shape of 3D objects (Ramachandran, 1988; Todd et al., 1997; Todd et al., 2023; Liu and Todd, 2004; Todd, 2004; Todd et al., 2014; Norman et al., 2004). The brain also uses binary disparity (Gonzalez and Perez, 1998; Murphy et al., 2016; Mitchison and Westheimer, 1984), motion parallax (Kim et al., 2016; Yoonessi and Baker, 2011), and textures (Ichihara et al., 2007; Thomas G. et al., 2002; O’Brien and Johnston, 2000) for depth perception, which are most likely involved in the perception of stereoscopic objects. In addition, the interaction effect among disparity, motion parallax, and shading was observed in a previous study (Schiller et al., 2011), indicating that the behavior of these factors is complicated, which makes it difficult to investigate their effects. Reaching performance can also be affected by irrelevant factors due to the added cognitive load of using virtual reality. It has been reported that simply adopting tasks where real movements are represented on a monitor, translated into a cursor on the screen, can induce a perceived delay that does not actually exist, even in non-immersive VR settings (Kim et al., 2020b, Kim et al., 2021). Considering this, the factors contributing to the floor effect in the older group observed in the present study remain elusive.

We considered movement over the interval using feedforward control in the analysis. Thus, the significant differences between the 2D and 3D objects indicate that stereoscopic information is involved in movement planning. When a reaching task performed in a 3D VR environment requires depth perception for successful performance, an extended Fitts’ law model that adds terms related to depth information to the traditional Fitts’ law model better predicts movements (Clark et al., 2020), which is associated with the relation between difficulty and target depth. The stereoscopic effect in this study may have served as an additional requirement in depth movement. Because a better performance was shown with 2D objects, visuomotor transformation during movement planning should not include the stereoscopic effect for better performance. Moreover, the difference in the stereoscopic effect may be attributed to visuomotor adaptation. The participants in the present study controlled the cursor on the screen using the controller. For visuomotor transformation, the brain must find an appropriate calibration model that matches the actual arm position to the cursor position in the VR space. In a previous study, a differently mapped cursor induced by visuomotor adaptation influenced depth perception (Volcic et al., 2013). Alterations in internal calibration caused by this classical visuomotor adaptation task may be intrinsically similar to the differences triggered by adding the stereoscopic effect. In addition, brain activities related to learning for visuomotor adaptation have been observed (Paz et al., 2003), supporting a neural basis. Another previous study reported that the brain may perform calibration suitable for binocular vision when monocular vision is input, leading to an underestimation of the distance in a reaching task (Servos, 2000). These incorrect internal calibration models may have been influenced by the stereoscopic effect.

What factors contributed to the age-related differences? Let’s revisit our aim and task. We aimed to observe behavior changes influenced by stereoscopic effects through a ballistic reaching task, which involves multiple cognitive processes. First, depth perception itself requires additional cortical processing, particularly in the visual cortex (Backus et al., 2001; Thomas O. M. et al., 2002; Smith and Parker, 2021). This processing becomes more demanding for older adults due to general declines in neural efficiency and adaptability. Furthermore, changes in binocular vision, such as near exophoria and reduced convergence ability, are common with aging (Leat et al., 2013), and studies have shown that binocular vision processing is altered in older adults (Yan et al., 2021). Such disturbances can significantly impact tracking performance, as seen in manual tracking tasks where performance is poorer under monocular vision compared to binocular (Kim et al., 2023). Aging is also associated with a decline in depth perception from binocular disparity (Norman et al., 2000; 2006). Collectively, these changes in visual processing, particularly in stereoscopic depth perception, contribute to the observed age-related differences in reaching performance. Beyond depth processing, aging also affects motor behavior through changes in sensory integration and control strategies. A critical factor is the increased reliance on visual feedback for motor tasks. Older adults tend to rely more on visual feedback (Seidler et al., 2010), which compensates for a decline in proprioceptive control (Boisgontier et al., 2012; Olszewska et al., 2016). Proprioceptive control is essential for effective feedforward control (Wolpert et al., 1995), enabling quick online corrections during reaching (Kasuga et al., 2022). Delayed responses in adjusting movements when the target position shifts suddenly (Sarlegna, 2006) suggest reduced use of proprioceptive input, which forces older adults to rely more on visual feedback for making corrections in reaching tasks. Given that our task emphasized feedforward control, similar to typical reaching actions in everyday life, these likely contributed to the observed differences in performance. Our findings on age-related differences in reaching performance align with previous research that observed a floor effect in tracking tasks, where older adults’ performance showed reduced sensitivity to depth-related visual information, likely due to age-induced constraints (Kim et al., 2024). Taken together, these findings indicate that changes in stereoscopic depth perception, combined with shifts in sensory integration and motor control strategies, could impact reaching performance in older adults. Our results suggest that the use of stereoscopic effects in VR-based tasks and interventions should be approached with caution, especially considering the age-related changes in visual and motor processing.

In this study, we confirmed the stereoscopic effect of objects in VR environments. However, we did not set sophisticated parameters to adjust the stereoscopic effect, as we initially aimed to determine a rough effect. We discovered the floor effect by age and the ceiling effect by distance. Additionally, movements through feedback control during ongoing reaching should be investigated in future studies. In this case, the effect of motion parallax could have affected the internal calibration. Naturally, users may subconsciously strive to estimate the distance between the objects and the virtual camera used for the first-person view implemented in the VR space. Thus, this distance should also be investigated in the future, including whether the perceived distance is identical to the actual distance between the target and the camera. Moreover, while we focused on two specific distances (10 cm and 20 cm) as proxies for task difficulty, future studies should investigate whether the stereoscopic effect generalizes across a broader range of distances. Furthermore, while we focused on young and older participants to capture distinct age-related differences, future studies could include middle-aged participants to explore potential regressive changes in movement planning or identify an inflection point in motor control across the lifespan. Additionally, future research should account for participants’ prior VR experience and systematically control IPD, as these factors may influence how participants interact with virtual environments. Including a group of experienced VR users could further provide valuable insights into performance differences across age groups. In conclusion, we confirmed that the effect of stereoscopic objects is not negligible in VR reaching tasks.