The study was designed as a within-subjects with four experimental conditions.

• Cycle only condition (A), in which the participant is cycling in front of a black screen;

This experimental design was conceived to explore the impact of an increasingly engaging experience on physical exercise responses and subjective experience. Condition A provides a traditional activity without additional stimuli. Condition B involves a limited level of visual engagement, which is made more immersive in condition C. Finally, the interaction factor is added in condition D.

The experiment foresaw a single session, during which all participants performed the four conditions subsequently. The exposure to each condition was randomized. This randomization ensures that each participant was exposed to the experimental conditions in a different order, thus minimizing the risk of order effects such as learning, fatigue, or familiarity with the task. By randomizing the order, we aim to isolate the effects of each condition, while controlling for any confounding influence of exposure sequence. No formal rest periods were scheduled between conditions. While this means that residual effects, such as cybersickness or fatigue, cannot be fully excluded, the randomized order of conditions was intended to minimize systematic bias across participants.

Eligible participants were healthy young adults 18–40 years old, in good health condition, as they self-reported no known medical conditions at the time of the experiment and confirmed their ability to engage in mild/moderate physical activity for 30–60 min without any limitations or impairments. Participants were recruited between 1st September 2021 and 31st July 2022. Participants were excluded if they could not maintain the sitting position long, had severe visual impairment preventing a correct visualization of the virtual environment, or had a history of significant motion sickness or epilepsy. The study has been approved by the National Research Council institutional review board, i.e., the Research Ethics and Integrity Committee, with the Ethical Clearance protocol nr. 0,052,420/2021 [EC EVEREST] of 22nd July 2021. All participants signed a written informed consent before participating in the study and consented to data treatment in compliance with GDPR.

The experimenter introduced the procedure and invited the participants to familiarize themselves with the setup. The height of the ergometer seat, the size of the HMD, and the chest strap were adjusted to fit with the individual anthropometric features. We then randomly generated the sequence of experimental conditions, gave the participants instructions on the experiment procedure, and introduced the questionnaires that would be administered at the end of each condition. The experiment consisted of five subsequent phases described in Figure 1.

The session started with 5 minutes of rest, during which the participant was asked to stay relaxed and quiet while sitting on the ergometer without cycling. Heart rate and breathing rate were measured to obtain baseline physiological values prior to the experimental conditions. After 5 minutes, the participant was asked to take two breaths with maximum inspiration/expiration by inhaling/exhaling the maximum amount of air possible. Such baseline assessment was followed by a warm-up phase to identify the exercise intensity. The exercise intensity, represented by the workload (WL) of the ergometer, is kept constant during the four exercise conditions. The exercise WL, expressed in Watt, is defined as the one eliciting a heart rate within 57% and 76% of the maximum value of the heart rate of the subject (HR_max). To ensure that the heart rate would not exceed the range for the subsequent 40 min of exercise, we considered a heart rate value equal to 65% to start the experiment. The value of HR_max is derived from the subject’s age according to Fox’s equation (Fox et al., 1971). The range of %HR_max corresponds to a light to moderate intensity, according to the definition provided by the American College of Sports Medicine (Garber et al., 2011). The participant was asked to maintain a cycling cadence of 60 RPM while the experimenter gradually increased the workload starting from 30 W until the HR reached the target value. In the following four phases, lasting 10 min each, the participant cycled in the four experimental conditions, trying to keep a cycling speed corresponding to 60 RPM. The session ended with a cool-down phase and a final 3 minutes of rest.

The experimental setup represented in Figure 2 is based on the Virtual Park system used by patients in our clinical studies (Colombo et al., 2024; Mondellini et al., 2023). Considering that they are the final target users of our investigation, such a choice would facilitate transferring our results to rehabilitation settings.

The equipment mainly included a stationary bike, a visualization device, a chest strap, and a VR application. The bike was the ergometer COSMED E100, a medical device that can be interfaced with third-party applications. The ergometer is connected via cable to a USB serial port for exchanging data with the VR application, namely, workload and cycling cadence. We used a Zephyr Bioharness 3.0 device that allows real-time recording of several physiological parameters, e.g., heart rate, respiratory rate, breathing waveform, core temperature, activity levels, and posture. The monitoring system comprises a Bluetooth module handling the data exchange and recording and an adjustable chest strap, available in three sizes and embedding different sensors. These include passive conductive ECG sensors, a pressure sensor pad for detecting breathing action, a thermistor, and a 3-axis accelerometer to estimate subject activity level and posture. The equipment also includes a joystick for the interaction with the VE (in condition D), which is connected through an Arduino DUE board programmed to send a signal indicating the position of the joystick among the three possible ones: neutral (joystick is in the center), left or right. As a visualization device, we used a 65″ Philips LCD TV monitor for TV condition and an HTC Vive first generation headset for HMD and HMD + interaction. Finally, we used a VR-ready laptop (Alienware 15 R4 with NVIDIA GeForce GTX 1080) to connect all the devices and display the VE. The VR application, developed in Unity, is organized into different modules, each implementing a specific functionality. The experimenter GUI module has two panels: one for creating the user profile and another for controlling the session. The first panel gathers user information, while the second manages device connections, experiment phases, and real-time physiological feedback. The virtual environment module consists of the virtual park and the UI panel shown to the participant. The 3D environment represents a park that can be explored by riding along a predefined circular path with a virtual bicycle, which is controlled by the user’s cycling movement. The path is flat with no hills and no abrupt changes of direction. For a more detailed description of the VE and its development, we refer the reader to our previous studies (Colombo et al., 2019; 2024). The UI panel is superimposed to the VE and shows textual information on the cycling cadence and the current phase of the experiment during rest, warm-up, and cool-down. The SteamVR plugin has also been imported to allow the application to be deployed for the HTC Vive. For condition D, we added an interactive task that can be performed using a joystick. The task is a simple Go/No-Go task, i.e., the user must act when the Go stimulus appears, while no action must be done when the No-Go stimulus is on. In our application, the stimuli are colored spheres appearing on both sides of the path at random intervals between 1 and 4 s. The participant has to “collect” the blue spheres while avoiding the red ones. To collect a sphere, they have to move the joystick to the right or left, depending on the sphere’s position relative to the path. The cycling module includes communication with the real ergometer and navigation within the VE through the movement of a virtual bicycle. The communication protocol allows retrieving the value of cycling cadence, which is used to control the movement in the VE, and changing the workload, during the first phase of the experiment in which the experimenter gradually increases the workload directly from the GUI of the Unity application. The navigation is controlled by the user him/herself: the cycle-ergometer cadence value expressed in revolutions per minute (RPM) is converted into the virtual speed value, corresponding to the visual representation’s flowing velocity. An empirical estimation has been made to identify the optimal conversion factor. As it happens in reality, if the rider stops pedaling, the VE continues to slide until the cycle-ergometer wheel stops rotating because of its inertia. The navigation occurs along a predefined circular route, which simulates a path in a park. The route has been created by placing subsequent nodes in predefined positions. Navigation occurs by interpolating these nodes in real-time through quaternion spherical interpolation (slerp). This module also communicates with the GUI module to visualize the updated parameters and control the workload during the warm-up phase. The physiological sensor module handles the communication with the Zephyr BioHarness chest strap, and the retrieval of the physiological data needed for controlling the experiment. To integrate the data in Unity at runtime, we developed two ad hoc libraries starting from the SDK provided by the producer. In such a way, the heart rate and breathing rate are integrated in the Unity application, can be visualized in the GUI, and are directly sent to the report module to save the session data. The report module generates the report file containing the data needed for the subsequent analysis. The file, in CSV format, is organized into four sections. The first includes the participant’s personal data (ID, age, weight, height) that the experimenter inserts in the GUI fields. The second part comprises raw data saved with a frequency of 1 Hz. The saved parameters are: heart rate, breathing rate, cycling cadence, workload, and a string identifier indicating the protocol phase (rest, warm-up, condition A, etc.). In the third part of the report file, the values of RPE for each condition are saved. The final part of the file includes the performance metrics of the additional task: number of collected and missed target items and number of errors).

A demographic questionnaire was administered to collect information about participants: gender, age, usual weight and height, and habits related to physical activity and tobacco consumption. Objective and subjective outcomes were measured to investigate the effects of the experimental conditions. Objective data were recorded continuously during each experimental session. At the same time, the subjective evaluation consisted of a short questionnaire, administered at the end of each condition, that was made specifically for this study to fit the experiment procedure. In addition, we administered a preference questionnaire at the end of the experimental session.

The IBM SPSS software v.29 was used for statistical analyses. First, we performed normality check to calculate if the scores were distributed on the normal curve. Descriptive analyses were performed for all objective and subjective data acquired. In particular, mean, standard deviation, minor, and major values were obtained for variables with normal distribution, and median and interquartile range for variables with non-normal distribution. One-way repeated measures ANOVA was performed to evaluate the differences in the scores of each variable in the four experimental conditions, except for the spatial presence scores where the comparison was made between three conditions (B-C-D). When the variable had a non-normal distribution, the Friedman test was performed to observe differences between conditions. The effect size ( η _p ²) for the ANOVA tests and the Kendall’s coefficient (W) for Friedman tests are reported for each variable of interest. Bonferroni adjustment was used in the post-hoc analysis. For the subjective variables enjoyment and attention, both measured with single items on a five-point Likert scale, we preliminarily verified the normality of the distributions. The results supported the use of parametric analyses. This choice is consistent with the established practice of considering Likert scales with >5 levels as interval approximations for group statistical analyses, particularly in within-subject designs with moderate sample sizes (Sullivan and Artino Jr, 2013; Norman, 2010). Finally, to control for type I error, we prespecified two families of variables: (a) exercise-related (HR, BR, CC, RPE) and (b) user-experience (enjoyment, self-perceived attention, flow, spatial presence, preference). Multiple comparisons corrections (Bonferroni) were applied to the post-hoc comparisons within each type of variable.

Concerning our primary outcome, our results indicate that virtual reality, disregarding the level of immersion and the addition of interaction, does not directly influence perceived effort measured by the RPE scale. Such a result seems in contrast to what we expected based on the scientific literature. Generally, when the procedure allows the participant to control the workload, individuals tend to work harder or longer while perceiving the same level of effort (Runswick et al., 2023). When the workload is controlled as in our case, e.g., increased to simulate hills, VR could help participants to perceive a lower exertion (Finnegan et al., 2023). Similarly, in a study by Matsangidou et al., participants rated their RPE lower when performing a strength exercise in VR than when performing the exercise without feedback. RPE was assessed every 60 s, and the work rate was constant and equal to 20% of each participant’s 1RM (i.e., the heaviest weight they were able to lit). The duration of the task was determined by the participant, and corresponded to time to exhaustion (Matsangidou et al., 2019). Huang and colleagues compared four conditions of exercise on a stationary bike: a PC desktop, a projection, an HMD, and without any screen. The workload was controlled and started at 30 W with a gradual increase every 1 min until the anaerobic threshold was reached. They found a significantly lower perceived exertion in the immersive - wearing the HMD - and semi-immersive - using a projector - condition with respect to the control condition - during which no feedback was displayed - (Huang et al., 2008). The presence of a virtual environment, displayed on a semi-immersive projection, was also associated with lower perceived effort in the study by Mestre et al., in which the effort feedback was controlled through a pad fixed on the rear wheel and the participants were asked to maintain a moderate exercise intensity level for an average duration of 15 min (Mestre et al., 2011a). Our study’s findings align with other studies in which the workload or the expected work rate was controlled and fixed to a given value. Monedero et al. did not observe any differences in RPE between cycling while playing a videogame and without feedback (Monedero et al., 2015). Participants cycled for 30 min at a matched workload of 55% of their peak power output, and RPE was assessed every 10 min. Born and colleagues examined the impact of three factors, one being the level of immersion of a VR exergame where the user is required to perform different full-body movements to get into different poses (e.g., squat) represented by a given shape appearing on the wall placed in front of him/her. The exercise intensity, represented by the number of walls and poses, was the same for all participants and across all conditions. No differences were observed in perceived exertion between using a screen monitor and an HMD (Born et al., 2019).

Although a direct influence on RPE was not found, our study indicates that coupling immersion and interaction might induce a discrepancy between actual and perceived effort. Participants perceived the same level of effort, i.e., RPE, during all experimental conditions, while the actual physical effort was higher in condition D, as indicated by the significantly higher breathing rate values. Studies suggest that BR is a strong marker of actual physical effort showing a strong association to the time course of RPE with a variety of experimental interventions and high sensitivity to different fatigue states (Nicolò et al., 2017). According to this definition, our findings may suggest that participants perceived a lower effort than the actual one. To the best of our knowledge, there are no studies in which VR has influenced physical effort expressed through BR. However, some studies with a different experimental design, and not measuring BR, observe a similar tendency. Czub and colleagues found an indirect influence of VR on perceived effort in their study. No change in the self-reported effort was observed. However, participants performed a greater number of repetitions of the exercise, i.e., greater actual effort, when the task was performed in VR than when it was performed without immersive feedback (Czub and Janeta, 2021). Focusing on interaction, the additional interactive task in our study seemed to have a positive influence on the physical effort and induced higher self-perceived attention. Although a direct comparison cannot be made, a study by Wender et al. found that a more difficult item detection task induces higher mental effort and, as a consequence, greater pain. Such an increase in the perceived pain, in turn, made the participants perform worse in the cycling exercise, showing lower exercise tolerance (Wender et al., 2022). In our study, the additional task did not negatively influence either the perception of effort not the attention. Although preliminary, BR values, which were significantly higher in the HMD + interaction condition, suggest that our participants did not perceive greater effort while they were actually doing it. This may indicate that our task was simple enough to catch participants’ attention, which was significantly higher than in other conditions, without stressing them too much. Future investigations should include other tasks characterized by different difficulty levels and mental demands to confirm or mediate such a finding.

Concerning the participants’ subjective experience, condition D was the most enjoyable, even if not significantly compared to condition C. This result aligns with findings from other research where participants rated exercise in fully immersive mode as more enjoyable than non-immersive mode, but only when the execution of a task was present (Plante et al., 2003). Condition D is also the one in which participants felt more concentrated, significantly more than other conditions. On the other hand, no difference emerges among the four conditions as regards the perception of the passage of time. This result is not consistent with findings from other similar studies (Pallavicini and Pepe, 2019). However, the flow is typically measured as a multidimensional experience, while in our study, only the perception of time passage was measured. HMD conditions (C and D) are the ones eliciting the highest sense of spatial presence. The effect of interaction within a virtual environment on spatial presence, and sense of presence in general, is still a relatively young topic. As reported by Riches and colleagues, interaction with elements (i.e., avatars) within the VR environment increases the sense of presence (Riches et al., 2019). Therefore, indications suggest that the level of interaction in an environment enhances the experienced presence. However, our experimental design involved interaction tied to a cognitive task somewhat disconnected from the virtual environment, and this may explain why conditions C and D do not differ in the perceived sense of presence. Nevertheless, in condition D, a greater variability of answers (i.e., a greater standard deviation) can be observed. Moreover, while plausibility and place illusion are important aspects of the sense of presence in VEs, due to time and protocol constraints we could not perform an extensive evaluation of these dimensions by focusing on spatial presence. Future studies could investigate these factors in more detail to further understand their impact on VR experience.

In both immersive conditions (C and D), the number of participants reporting no side effects was the same (76.47% of the total). In condition C, more people complained of excessive sweating (3 vs. one of “D″) and visual problems. Although sweating is traditionally classified in the nausea category, it emerged that this discomfort was due to the temperature inside the laboratory and to the heat of wearing the headset, therefore for different reasons from those usually identified in the literature (Kennedy et al., 1993). While no participants specifically mentioned the weight as a problem, this discomfort may also partly reflect the additional weight of the device, although newer, lighter headsets could help limit this issue. In condition D, no subject reported visual difficulties; unlike condition C, however, nausea and greater disorientation symptoms emerged. This could be because participants - who had to concentrate on the Go-No Go task - felt lower freedom of possibility and, therefore, lower accommodation by the vestibular system. As reported in Ramaseri et al., movement in a virtual environment should be realistic to match sensory expectations (Ramaseri Chandra et al., 2022). Results on the subjective experience go hand in hand with the preferences expressed by the participants during the final questionnaire. In fact, the most preferred condition was D, followed by the other immersive condition, i.e., C.

Our study is not without limitations. The main limitation is about the population, which is unbalanced in terms of gender, with a percentage of females lower than 30%. One study indicates that gender may influence user experience in VR, hypothesizing that gaming habits may favor males by better evaluating their performance and ability to act in VR (Sagnier et al., 2020). On the other hand, other studies reported no gender differences in sense of presence, performance, enjoyment, and interest (Khashe et al., 2018; Rangelova and Marsden, 2018). A study reports that females are generally more susceptible to cybersickness because the distance between the headset’s lenses cannot fit female users’ interpupillary distance (Stanney et al., 2020). In our study, all participants were invited to adjust such a distance before starting the training until they were satisfied with the visualization of the virtual environment. Future studies should aim for a more balanced and representative sample to enhance generalizability and contribute to a clearer understanding of gender influences on user experience and perceived effort during exergames. Moreover, in a future study we aim to increase the sample size; this would help provide more robust and reliable conclusions also reducing the risk of Type II error. Moreover, there are limitations and potential improvements regarding the design of the experiment and of the conditions. Our study lacks a screen-based condition without a HMD but with interaction. This absence makes it challenging to conclude whether immersion alone impacts exercise responses or if interaction alone is responsible for observed effects. By including a screen-based with interaction condition, future studies could provide clearer insights into the relative importance of immersion versus interaction in virtual reality-based exercise interventions. With respect to the experimental setup, participants in the Cycle only condition could see the reflection of their bodies, which may have influenced their experience and represents a potential limitation of this condition. Moreover, a newer version of HMD for fully immersive conditions could influence the overall experience and results.

Our protocol focused on assessing perceived exertion in a condition in which the workload, and in an indirect way the HR, was controlled. Therefore, we cannot completely exclude that controlling the HR had influenced BR too and future investigations, e.g., through different study design, are recommended. Moreover, a future study could include the possibility of measuring exercise tolerance, which is a key aspect for our target users. For this purpose, the design of the experiment could be modified in self-paced exercise bouts in which participants are encouraged to go as hard, i.e., by directly controlling the workload through a button, as they could. Another potential limitation of our study design could be the cycling prescription, i.e., short bouts at a moderate intensity and cycling speed around 60 RPM may be too easy for this sample and not sufficiently challenging to enable changes in performance across conditions. Finally, our interaction task was a simple Go-no-go task with a single difficulty level and one type of stimuli (red or blue sphere) that did not change across the 10-min session. Higher complexity in terms of visual representation and characteristics of the task could enhance the effect on perceived effort and subjective experience and, therefore, deserves further investigation.

There are also some limitations that can be overcome in future studies with regard to measurements and statistical analysis. Single-item measures were used for flow and presence to minimize interruptions during the exercise bouts. Future studies could consider using multi-item validated scales to improve reliability and construct validity. Moreover, our study lacks a formal validity assessment for subjective measurements, such as self-perceived attention and flow, by participants who could have provided an additional indicator of the perceived clarity and relevance of the single item used. Future studies could include an evaluation of validity by asking participants to express their opinions on the clarity and relevance of the item, in order to enhance the understanding and reliability of the short questionnaire. Lastly, our study was exploratory and an a priori power analysis was not conducted. With N = 34, α = 0.05 and power = 0.80, for the main outcome (repeated measures ANOVA with four conditions), the study allows us to detect a minimal effect of Cohen’s f = 0.60, corresponding to a very large effect size (Cohen, 1988). This implies that the sample was adequate to detect large differences, while small to medium effects may not have been detected (Faul et al., 2007).

Finally, the population included in our study represents a limited part of those who could benefit from modification of perception of effort during exercise. The characteristics of the users, e.g., their age and health status, could influence the experience and, therefore, the effects of VR on perception. Indeed, although several studies have explored the acceptability and attitudes of older adults towards cycling in virtual reality, yielding positive results comparable to those of younger individuals, it is crucial to consider the influence of age on their experience and engagement Ortet et al. (2022); Tegegne et al. (2025); Hibbs et al. (2024). The way older adults engage and respond to virtual cycling experiences, as factors such as cognitive decline, physical limitations, and varying levels of familiarity with technology may affect their comfort, enjoyment, and overall perception of activity in virtual environments. One of our final goals is to transfer the results to the rehabilitation field. To do so, the same study protocol should be replicated with older adults with respiratory diseases. A future study with patients should take into account other potential influencing factors that may alter the results in terms of perceived exertion. The most critical one is the fact that patients with respiratory diseases may experience breathlessness at even low level of physical activity intensity.