Looking down with the head up: improving walking in VR and VST-AR with increased downward field of view

1 Introduction

To explore expansive virtual worlds, users rely on head-mounted displays (HMDs) with wide Field of View (FOV) and walk around the world. Numerous studies have aimed to enhance the FOV of HMDs, achieving significant improvements, particularly in the horizontal direction (Ratcliff et al., 2020; Rakkolainen et al., 2017; Xiao and Benko, 2016; Yamada and Manabe, 2016). However, it is crucial to note that vertical FOV plays a vital role in perceiving one’s position and movements within the virtual environment, including the space near the body and beneath the feet. Despite the advancements in horizontal FOV, efforts to expand the vertical FOV of HMDs have been limited (Endo et al., 2021; Nakano et al., 2021; Nakano et al., 2023; Nakano and Narumi, 2025; Xiao et al., 2025).

When walking, humans need to avoid collisions with objects to avoid injury and tread the ground accurately to avoid losing their footing. However, the limited downward FOV of HMDs can make it difficult to visualize virtual objects and ground near the body without significantly turning the head downward in a virtual environment (VE). This problem may also represent a potential factor contributing to increased fear. An expanded downward FOV is important for walking in virtual environments, where users must navigate around objects and step onto the ground while maintaining a forward-facing position. Nakano et al. developed an HMD with an enlarged downward FOV and demonstrated that, while exploring a VE using the controller’s trackpad, the sense of presence [the feeling of being in the VE (Sanchez-Vives and Slater, 2005)] improved, the head angle turned upward, and users’ ability to estimate their current position increased (Nakano et al., 2021). However, it remains unclear how users perceive virtual objects close to the body and how their physical behavior changes when avoiding or stepping on them. Therefore, we focus on walking within a VE to examine whether an expanded downward FOV prompts users to tilt their heads upward while walking and enhances interactions with virtual objects using their feet.

The issue of limited downward FOV is not exclusive to Virtual Reality (VR) but also affects Augmented Reality (AR). HMDs designed for AR typically feature narrower FOVs compared to those for VR, and unique AR challenges such as virtual object misalignment and image latency persist. While advancements such as Video See-Through (VST) HMDs, exemplified by Varjo XR-4, Apple Vision Pro, and Boba 3 (Xiao et al., 2025) have been developed using VR-HMD technology to offer wider horizontal FOVs, the exploration of peripersonal space in AR remains constrained due to the inherent limitation of downward FOV.

Several studies (Horsak et al., 2021; Canessa et al., 2019; Brument et al., 2019; Knight and Baber, 2007) have pointed out that these FOV limitations are also problematic when comparing walking in VR with walking in a real environment. Although walking characteristics in VST-AR are not expected to be the same as those in the real environment due to problems with FOV, latency, and stereopsis, it is worth noting that many features can be shared with VR and are easily compared. As VST-AR HMDs become more popular, it is important to investigate walking characteristics in VST-AR, since many users will walk around town using AR. In contrast, human perception and resulting behavior may differ between VST-AR, which presents camera-captured real-world imagery to create a realistic visual experience, and VE, which is composed of virtual polygons. Previous studies have reported that expanding the downward FOV enables users to perform tasks while keeping their heads oriented forward, improves sense of self-location (Nakano et al., 2021), and facilitates reaching movements with the hands and feet (Nakano and Narumi, 2025). These effects have been attributed to increased realism in VEs, which may allow users to perform more natural body movements.

In this study, we examine whether the effects observed in VR, resulting from an expanded downward FOV on presence, walking, and interaction, also manifest in VST-AR, characterized by a narrower FOV, using a VST-HMD with an increased downward FOV (Nakano et al., 2023). By comparing identical tasks performed in VR and VST-AR, we also examine whether the observed benefits of downward FOV expansion may be associated with increased realism.

The main contributions of this study are outlined as follows:

• The expanded downward FOV enables users to walk and observe virtual objects on the ground without needing to tilt their heads downward, making it easier to avoid or step over virtual objects.
• The widened downward FOV facilitates users in perceiving and interacting with virtual objects within their immediate surroundings, notably increasing leg movement when avoiding these objects. Exploratory analyses further suggest that it may enhance the realism and the perceived three-dimensionality of virtual objects.
• The effects of the expanded downward FOV were observed in both VR and VST-AR, although the effects were weaker in VST-AR. Participants showed larger foot movements when avoiding virtual objects in VR than in VST-AR.

2 Related work

2.1 HMD with wide FOV

HMDs with wide FOVs are well known for enhancing immersive VR experiences, notwithstanding the potential risk of inducing cybersickness (Lin et al., 2002). However, traditional wide-FOV HMDs primarily concentrate on frontal or wide horizontal image rendering (Ratcliff et al., 2020; Rakkolainen et al., 2017), neglecting content display in the downward FOV. This necessitates significant downward head movements to view such content. Efforts are underway to expand HMD viewing angles by leveraging human visual characteristics, which exhibit reduced sensitivity in the peripheral vision compared to the central visual field (Ishiguro and Rekimoto, 2011). Inexpensive LED arrays (Xiao and Benko, 2016) or Fresnel lenses (Yamada and Manabe, 2016) were added to the main optics to enhance horizontal and vertical viewing angles, albeit at a lower resolution. An HMD that allows users to experience a VE while viewing the surrounding physical environment from a downward FOV by adding an adjustable transparency liquid-crystal shutter to the lower part of the HMD has also been studied (Lindeman, 2016; Tran et al., 2019); however, it does not support video presentation. In contrast, Tanikawa et al. successfully manipulated walking direction by displaying optical flow on a smartphone attached to the user’s cheek (Tanikawa et al., 2012), suggesting that video display for peripheral and downward FOV is effective.

To present high-resolution images in the downward FOV, research is being conducted on HMDs with new sets of lenses and LCDs added to the sides and bottoms of the devices. Endo et al. integrated a smartphone into the sides and bottom of an HMD to achieve wide horizontal and downward FOVs (Endo et al., 2021). Nakano et al. expanded the downward FOV further by incorporating a Fresnel lens and an LCD into the bottom of an HMD (Nakano et al., 2021). Furthermore, they enabled a wide downward FOV in VST-AR by attaching a fisheye camera with a wide FOV in front of an HMD (Nakano et al., 2023). Therefore, the HMD used in this study, which integrates VR and VST-AR for walking and interacting with virtual objects, follows the system proposed by Nakano et al. (Nakano et al., 2023). It boasts a wide downward FOV and is capable of displaying VST-AR.

2.2 Walking with wide downward FOV

Walking is a common real-world human movement to relocate from one place to another. It is also viewed as a natural and promising method for VEs (Nilsson et al., 2018b). However, limitations in the downward FOV impede users’ walking capabilities, for instance, when navigating uneven or slippery terrain (Marigold and Patla, 2008). Several studies (Authié et al., 2017; Turano et al., 2001) in neuroscience have shown that when the human visual field is restricted, the gaze turns downward during walking, supporting the importance of the downward FOV. The downward head movement is due to the ground surface providing important information for controlling posture and providing a stable reference (Flückiger and Baumberger, 1988; Bian et al., 2006; Kluzik et al., 2007). It has also been shown that a reduced viewing angle causes the head to move more to compensate for reduced visibility, resulting in reduced postural stability (Venturino and Wells, 1990; Knight and Baber, 2007; Taneda et al., 2021).

A limited FOV increases the time required to reach the destination and increases the likelihood that the user will encounter obstacles in a VE (Hassan et al., 2007). Jones et al. reported that limiting the vertical FOV reduces the ability to judge distances in a VE (Jones et al., 2016). In addition, in real environments, walking speed decreases when the downward FOV is limited, especially when descending stairs, leading to increased head tilting downward (Miyasike-daSilva et al., 2019). In a VE with reduced feet visibility due to a limited downward FOV, users may experience heightened fear, difficulty walking, and excessive downward head tilting.

Walking is considered one of the most challenging forms of locomotion in VEs, due to the spatial constraints imposed by virtual environments compared to real ones. Strategies such as Walking in Place (Nilsson et al., 2016), which mimics movement by stepping in place, and Redirected Walking (Razzaque et al., 2001; Nilsson et al., 2018a), which alters the consequences of the user’s physical translation/rotation, have been proposed and are actively researched to address this issue. Various studies aim to enhance the effectiveness of Redirected Walking; for horizontal FOV (Williams and Peck, 2019) and velocity increases (Williams-Sanders et al., 2019), head angle manipulation (Marchal et al., 2010), and hill walking speed manipulation (Matsumoto et al., 2017) have shown effectiveness for vertical Redirected Walking. However, peripheral visual information is important for perceiving speed through foot movement and changes in the ground. Thus, limitations in downward FOV may diminish the efficacy of traditional Walking in Place and redirection methods reliant on visual manipulation.

2.3 Importance of downward FOV in AR

The limitation of the downward FOV affects not only VR but also systems and research that use AR-HMDs to superimpose information on the real environment. For example, when information is presented in the space near the body using a VST-HMD, the head must be turned downward significantly due to the limited downward FOV. When paramedics perform medical procedures, the frequency of head bending, rotation, and side bending increases in HMD-VR compared to the real environment (Knight and Baber, 2007), and the same problem is expected to occur in AR. It has also been noted that the downward FOV limitation reduces visibility near the mouth, potentially limiting the applicability of the gustatory manipulation effect that alters the food’s appearance (Nakano et al., 2023). We argue that verifying whether the downward FOV is also effective in VST-AR will improve its applicability.

It is also essential to verify whether the effect of downward field expansion differs between VR and AR. Jones et al. found that egocentric depth perception in VR differs from that in real and AR environments (Jones et al., 2008). Studies on spacecraft operations reveal that VR performance closely resembles real-world performance and surpasses that of AR and hybrid reality setups (Banerjee et al., 2022). Moreover, participants in immersive projection environments tend to avoid virtual objects more than real ones (Sanz et al., 2015), and a mismatch between the user’s self-avatar foot’s movements and the VE can lead to altered behavior (Pan and Steed, 2019). During walking, the VE’s limited viewing angle compared to the real environment led to reduced speed and instability, and to large head movements (Horsak et al., 2021; Canessa et al., 2019; Brument et al., 2019). Note that these results may also be influenced by the VR and AR HMDs themselves (and their calibration). While the expansion of downward FOV is anticipated to yield positive effects with real-world applications, its impact may differ across VR, AR, and real environments, necessitating further investigation. Further, although no consistent trend linking increasing cybersickness with expanding the downward FOV has been identified (Nakano et al., 2021; Nakano and Narumi, 2025), it warrants measurement, as some studies suggest a correlation between increasing FOV and cybersickness (Xiao and Benko, 2016). Therefore, this study investigates the effects of expanding the downward FOV on behaviors such as walking, head orientation, and cybersickness from both VR and VST-AR perspectives. Experimental tasks involved avoiding and stepping on virtual objects placed on the floor to measure behavioral characteristics resembling common walking behaviors.

3 HMD that expands the downward FOV

In this experiment, we used HMDs capable of presenting images in a downward FOV in VR and VST-AR (see Figure 1). This HMD prototype builds upon previous studies on HMDs with wider downward FOVs in both VR (Nakano et al., 2021) and VST-AR (Nakano et al., 2023). To expand the FOV downward, we modified the HTC VIVE by removing its external components. We integrated two sets of LCD modules (Sharp LS029B3SX02) with dimensions of 2.9 [inches] and a display area of 51.84 × 51.84 [mm], featuring a resolution of 1440 × 1440 [pixels] and frame rates of 120 [fps]. Additionally, Fresnel lenses with a length of 35 [mm], a width of 38 [mm], and a focal length of 40 [mm] were positioned beneath each eyepiece at a 20 [deg] angle. The eye relief and the distance from the lens to the LCD were both 28 [mm], which is shorter than the lens’s focal length, but the image was observed clearly. The distance between the user’s eyes and the HTC VIVE eyepiece was adjusted to approximately 13 [mm] at its maximum extent to accommodate the additional display units. For VST-AR visual presentation, a fisheye camera (ELP-SUSB1080P01-L170) with a resolution of 1080 × 1920 [pixels], a FOV of 170 [deg] and a frame rate of 50 [fps] was added to the front of the HMD prototype. The camera was tilted downward by 25 [deg] from the front of the HMD. The measured system delay was approximately 200 [ms]. Further, our HMD’s weight, excluding the cord, is approximately 715 [g], an increase of approximately 200 [g] from the original.

The actual FOVs were measured using a calibrated equidistant projection fisheye camera (GS-15WDCM-1.5 MM) with a resolution of 1920 × 1080 [pixels] and a FOV of 180 [deg]. The results are presented in Table 1. As indicated in the tables, the implemented prototype HMD maintains the same viewing angle as in the previous VR study (Nakano et al., 2021; Nakano and Narumi, 2025). However, the vertical viewing angle for the VST-AR is 10 [deg] less than in the previous study (Nakano et al., 2023). The quad size for displaying images from the fisheye camera placed in front was reduced by 20 % compared to parameters used in a previous study (Nakano et al., 2023). This adjustment accommodates participants’ standing and walking, requiring a greater focal distance than in the previous study, which focused on visualizing food near the mouth. The three-dimensional (3D)-printed enclosure reduces the original HTC VIVE’s downward FOV by approximately 10 [deg]. As depicted in the figure, this gap appears to split the images, with some users likening it to the frame of a pair of glasses. To mitigate future decreases in FOV, a dedicated fused lens could be employed.

TABLE 1. The FOVs of the VR and VST-AR were assessed using a calibrated equidistant projection fisheye camera.

The original HTC VIVE had a diagonal FOV (VR) of approximately 75 [deg]. Please note that the FOV is presented as a range because a portion of the VST image is not displayed in the added downward viewing area.

Figures 1D,E shows the images on the actual HMD with an increased downward FOV. The original FOV of the HTC VIVE, illustrated in blue, is insufficient to perceive downward-placed additive objects and virtual shoes. To view them, the head must be turned significantly downward. Our HMD design with an expanded downward FOV can present images with the added downward FOV shown in orange, in addition to the original FOV of the HTC VIVE, as shown in Figure 1D, allowing the viewer to see virtual objects and shoes while looking at the front view. Figure 1E shows an image provided by VST-AR with an increased downward FOV. Viewing virtual objects and shoes while facing forward is possible, as in VR. In this experiment, to enhance participants’ perception of their own feet positions, virtual objects and shoes were displayed without restrictions, even if they fell outside the FOV of the VST-AR. The fisheye camera used in this system is monocular; therefore, participants cannot perceive stereoscopic depth in the VST-AR. In contrast, virtual objects and shoes are rendered stereoscopically. This difference may lead to different depth impressions when comparing VST-AR and VR, potentially influencing the experimental results.

The reduction in the vertical viewing angle in VST-AR compared to VR is attributed to the distortion-corrected fisheye camera used in this study. Despite having a 170 [deg] viewing angle, it falls short of fully covering the vertical angle required for a downward-focused HMD. Using a fisheye camera with a wider viewing angle or integrating multiple webcams might be imperative. The software for the proposed system was developed using Unity 2019.4.39f1. Within Unity, two virtual cameras were utilized to project images onto the LCDs, precisely aligned with the main camera of SteamVR. To ensure accurate image presentation, the initial intrinsic and extrinsic parameters of the virtual cameras were meticulously determined, taking into account the physical dimensions of the display units. The edited view frustum was leveraged to rectify the image output from the virtual sub-camera. Furthermore, the distortion introduced by the Fresnel lens was manually corrected through mesh deformation.

4 User study

4.1 Overview

The study aimed to assess whether employing an HMD with an expanded downward FOV enhances the visibility of virtual objects and feet, as well as the sense of presence in the VE and virtual objects. Foot movements were tracked while stepping over and stepping onto virtual objects to examine whether the downward FOV influences body movements and whether the effect of a widened downward FOV is evident in both VR and VST-AR environments. The experimental setup included a prototype HMD, three optical motion-capture cameras (VICON MX 13), motion-capture software (VICON Tracker 3.9.0), and position markers affixed to participants’ feet. Participants traversed a 3 (m) experimental course and engaged in tasks involving avoiding or stepping on a virtual object positioned at an intermediate point along the course (see Figure 2).

The experiment followed a 2 × 2 factorial design with independent variables, including (1) the presence of an expanded downward FOV (ON: utilizing an additional display unit, OFF: without utilizing it) and (2) the environmental appearance (VR: walking in a VR environment; AR: walking in a VST-AR environment). Consequently, four conditions were examined (OnA: On_AR; OnV: On_VR; OffA: Off_AR; and OffV: Off_VR), all within-subject variables. For these variables, two types of interaction tasks with the virtual object were used (Avoid: avoided by stepping over the virtual object, StepOn: Step onto the center of the virtual object). Participants performed all four tasks in a counterbalanced order based on the Latin Square design, either in the Avoid or StepOn condition, and then in the other condition in the same order for a total of eight different tasks. In the experiment, participants walked a 3 [m] experimental path and performed the task of stepping over (Avoid)/onto (StepOn) a virtual object placed at the midpoint.

Participants were instructed to step over the Avoid Object [W: 30 × D: 120 × H: 30 (cm)] in the Avoid condition so that their feet did not hit it and step onto the red area [W: 10 × D: 10 × H: 3 (cm)] in the center of the Step Object [W: 30 × D: 30 × H: 3 (cm)] in the StepOn condition. The virtual shoes used in the experiment are depicted in Figure 2C and were synchronized with position markers attached to participants’ feet. This visual aid helped participants locate their feet and estimate the distance to the virtual object. Additionally, a small white cube served as a marker, and participants were instructed to step onto this marker with their feet to hit the center of the step object during the StepOn task.

4.2 Procedure

Participants were asked to indicate their confidence in their dominant foot prior to commencing the experiment. When the dominant foot was uncertain, participants were prompted to demonstrate which foot they would use to kick an empty box in a real-world scenario. We referenced a relevant study (van Melick et al., 2017) for determination. Initially, participants donned position markers on both feet and an HMD, then adjusted the position of the virtual shoe and foot under the OnA condition. Subsequently, a practice task involving two environmental conditions (VR or AR) and two types of interactions with virtual objects (StepOn or Avoid) was completed four times, with the downward FOV enlarged. This practice task aimed to acquaint participants with downward FOV image presentation and interaction with virtual objects. Following a 3 [min] break, during which participants reviewed the questionnaire content to be utilized in the experiment with the HMD removed, the following procedures were executed under each condition.

• Participants wore the HMD and trackers.
• They moved to the starting point and faced the front.
• They started walking.
• They interacted with the virtual object (step/avoid) and moved to the goal.
• They returned to their starting position.
• Step 2 – 4 were repeated four times.
• They removed the HMD and completed the questionnaire.
• They took a 3 [min] break.

The questionnaire administered included the I-group Presence Questionnaire (IPQ) (Schubert et al., 2001) to evaluate the sense of presence in VR. Additionally, the Augmented Reality Presence Questionnaire (ARPQ) (Regenbrecht and Schubert, 2021), adapted for this experiment using the questionnaire proposed by Regenbrecht and Schubert, was utilized to gauge the presence of virtual objects in AR. The Simulator Sickness Questionnaire (SSQ) (Kennedy et al., 1993) was employed to assess the impact of an increased FOV on cybersickness. The questionnaire also included three author-developed questions about walking (ease, fear, and avoidance/stepping), along with a section for free-text feedback. For objective measurements, the following metrics were recorded: total 2 [m] transit time centered on the virtual object (Time), head movement pattern (Head Pitch Angle), and the trajectory of the participant’s dominant foot movement and distance to the virtual object (Foot Distance).

4.3 Measurements

Participants’ head movements were tracked using an HMD, while foot movements were measured using position markers affixed to their feet. Head and foot trajectory data were measured at a sample rate of 90 [Hz] when the user’s head was within 3 [m] of the experimental course. To facilitate the analysis of foot trajectories during participants’ interaction with the virtual object, the measured data were filtered within a 45 [cm] range in front of and behind the virtual object and within a 90 [cm] height range. This 45 [cm] range corresponds to three times the object depth (30 [cm]), i.e., the object size plus equal margins in front of and behind the object (30 + 30 + 30 [cm]). During the Avoid task, participants were instructed to continue moving forward without halting after stepping over the virtual object with their dominant foot. To ensure uniformity in foot movements among participants while avoiding the virtual object, participants were directed to vertically raise their thighs without lateral leg movement. The distance between the participant’s dominant foot and the virtual object was calculated as the nearest neighbor value in two dimensions, excluding the lateral direction (depth) from the participant’s perspective during the Avoid task. Additionally, the distance calculated in 3D was utilized when the virtual object was initially touched during the StepOn task. The decision boundary for contact with the virtual object was expanded by 3 [cm] from the Step Object to the top (W: 30 × D: 30 × H: 3 + 3 [cm]) to ensure contact with the foot position markers attached to the instep of the foot in the StepOn task.

4.4 Questionnaire

The IPQ was used to evaluate the presence of VR. In this experiment, four of the subitems of the IPQ (Pres: General Presence; SP: Space Presence; Inv: Involvement; and Real: Experienced Realism) were rated using a seven-point Likert scale [1 (strongly disagree) – 7 (strongly agree)]. Pres, SP, Inv, and Real consisted of one, five, four, and five items, respectively. The ARPQ was used to evaluate the presence of AR. Seven items were used in this experiment and measured on a seven-point categorical scale [1–7]. The meaning of each item is shown in Table 2. The SSQ was used to evaluate cybersickness. In this experiment, among the subitems of the SSQ, three (N: Nausea; O: Oculomotor; and D: Disorientation) were evaluated using a four-point scale [0 (None) – 3 (Severe)] and weighted for each value using Kennedy’s method (Kennedy et al., 1993). Each index consisted of seven questions. In addition, Total Severity (TS), the sum of each item’s values, was used as the evaluation index. Each value was weighted by the corresponding value. The author-developed questions were used to assess ease and fear of walking, as well as the task’s ease. Three items were arranged for this experiment and measured using the seven-point Likert scale [1 (strongly disagree) – 7 (strongly agree)]. The meaning of each item is shown in Table 2.

TABLE 2. Augmented Reality Presence Questionnaire (ARPQ) (Regenbrecht and Schubert, 2021) and author-developed questionnaire.

4.5 Participants

Participants were recruited via campus mailing lists and announcements. Following the guidelines of the Ethical Review Committee for Research Involving Human Subjects, Nara Institute of Science and Technology (2020-I-17), informed consent was obtained from each participant after a detailed explanation of the study. The experiment lasted approximately 2 hours, and each participant received compensation equivalent to about 20 USD. Initially, 18 participants were recruited for the study. However, data from two participants who did not adhere to the correct experimental procedures were excluded from the analysis. One participant did not check part of the questionnaire before the experiment, and the other went straight ahead without avoiding the object in the Avoid condition; consequently, the experimenter stopped the experiment midway and treated the data as error data. Therefore, the data of 16 participants (eight males and eight females) with a mean age of 23.6 years (standard deviation = 1.27) were used for analysis. Among these 16 participants, 15 reported their right foot as dominant, and one reported their left foot as dominant.

4.6 Hypotheses

In this experiment, we set the following hypotheses:

Previous studies have shown that expanding the downward FOV in VEs increases presence (Nakano et al., 2021; Nakano and Narumi, 2025). This study examines whether the same effect occurs during walking and in VST-AR environments.

• H1 Increasing the downward FOV increases presence, which in turn increases the IPQ and ARPQ scores.

When the downward FOV is restricted, people tend to look downward and increase head movements to obtain visual information necessary for walking (Authié et al., 2017; Turano et al., 2001; Venturino and Wells, 1990; Knight and Baber, 2007; Taneda et al., 2021). Therefore, expanding the downward FOV may allow users to walk and interact with obstacles more easily while keeping their heads oriented forward.

• H2 Increasing the downward FOV makes walking seem less fearful and increases the ease of walking and speed.
• H3 Increasing the downward FOV makes it easier to avoid an object with minimal movement or to accurately step on it.
• H4 Increasing the downward FOV lowers the average downward pitch angle of the head (the head turns upward) and causes less variance in the head motion.

The limitation of the downward FOV is also a problem in VST-AR, and expanding it is expected to improve walking, as in VR. However, because VST-AR images are not stereoscopic, include latency, and have a narrower FOV than VR, the effect is expected to be smaller in AR.

• H5 The effects of increasing the downward FOV on presence, walking, and interaction with virtual objects occur not only in VR but also in VST-AR; however, they are expected to be weaker in VST-AR than in VR.

5 Analysis and results

5.1 Overall

Table 3 displays the Wilcoxon signed-rank test results for IPQ and ARPQ, including Z-values, p-values, and Effect size r. IPQ specifically targeted the VR condition, while ARPQ focused on the VST-AR condition. Consequently, a Wilcoxon signed-rank test was conducted for the VR condition (OnV and OffV) and the VST-AR condition (OnA and OffA) to compare the presence in VR or VST-AR condition with and without the additional LCD units (ON vs. OFF). Table 4 presents the two-way ANOVA results, including F-values, p-values, and Effect size η p 2. This analysis considered the presence or absence of additional LCD units (ON vs. OFF) and environmental appearance (VR vs. AR). Given that the data did not adhere to a normal distribution, an aligned rank transform was employed for hypothesis testing (Wobbrock et al., 2011). To control the false discovery rate, p-values were adjusted for multiple comparisons using the Benjamini–Hochberg procedure with q = 0.05. Each hypothesis was treated as a separate correction family: the 22 Wilcoxon tests for H1 in Table 3 form one family, and in Table 4 the LCD main effects are grouped per hypothesis (H2: n = 6; H3: n = 5; H4: n = 4), while the App main effect and the App:LCD interaction each form a family of the 15 tests used to evaluate H5. SSQ items were not part of hypothesis testing and were not included in any correction family. Post hoc tests with multiple-comparison correction were planned if the interaction was significant; however, none reached significance, so none were performed.

TABLE 3. Experimental results of Z and p-values, along with effect sizes for IPQ (OnV vs. OffV) and ARPQ (OnA vs. OffA) in the Avoid condition and the StepOn condition.

The p-values are adjusted for multiple comparisons using the Benjamini–Hochberg procedure (FDR, q = 0.05). The family is defined as the 22 tests used to evaluate H1.

TABLE 4. Experimental results of F and p-values and effect sizes under Avoid conditions and StepOn conditions.

The Environment condition is abbreviated as “App”. The downward FOV condition is abbreviated as “LCD”. “App:LCD” indicates interaction. The p-values are adjusted for multiple comparisons using the Benjamini–Hochberg procedure (FDR, q = 0.05). For the LCD column, the family is the set of tests used to evaluate each hypothesis: H2 (n = 6; Easy, Fear, Time rows), H3 (n = 5; Avoid/Step, Foot Distance, and Landing Position rows), and H4 (n = 4; Head rows). For the App column and the App:LCD column, the family is the 15 tests used to evaluate H5 (App main effect and App:LCD interaction, respectively). SSQ items are not part of hypothesis testing, and their p-values are shown without correction. Items significant after correction are highlighted in bold. F values marked with * differ from the other F values.

Notable results for each condition are illustrated in (Avoid: Figures 3–8) and (StepOn: Figures 9–11). In the figures, the red crosses denote the average values. Significant differences are indicated by symbols (* for p < 0.001, for p < 0.01, and * for p < 0.05). These box plots show the interquartile range; the red central line indicates the median, and the whiskers extend to the minimum and maximum values; values beyond these limits are shown as outliers.

5.2 Results of avoid condition

Figures 3, 4 display the results of the IPQ and ARPQ, which assessed the presence of VR and AR, respectively. After correction, the Wilcoxon signed-rank tests did not reveal a significant effect of LCDs (ON vs. OFF) on any IPQ subscale or ARPQ item. Figure 5 shows the results of the author-developed questionnaire, which investigated the perceived ease of walking, the fear of walking, and the ease of avoiding. Two-way ANOVA revealed a significant main effect of LCDs (ON vs. OFF) on Avoid (M e a n ± S D, ON: 5.09 ± 1.63, OFF: 3.22 ± 1.64, p < 0.001). No significant interaction was found.

Figure 8 displays the results of the head pitch angle. Two-way ANOVA revealed a significant main effect of LCDs (ON vs. OFF) on Average (ON: 12.42 ± 11.32, OFF: 26.01 ± 13.57, p < 0.001) and Variance (ON: 8.52 ± 6.90, OFF: 23.33 ± 14.79, p < 0.001). No significant main effect of environment appearance (VR vs. AR) and interaction were found. Figure 6 (Left) illustrates the distance between the participant’s dominant foot and the virtual object when it was closest to the participant. The data were averaged over four trials for each participant under each condition to obtain a single set of data. Figure 7 shows the trajectory of the dominant foot when the participant avoided the virtual object. As the bias in the landing position of the dominant foot when avoiding the virtual object could be read in Figure 7, it was tabulated as Landing Position, analyzed and illustrated in Table 4 (Avoid) and Figure 6 (Right).

Two-way ANOVA revealed a significant main effect of LCDs (ON vs. OFF) on Foot Distance: nearest neighbor value (ON: 0.41 ± 0.07, OFF: 0.33 ± 0.09, p < 0.001) and Landing Position (ON: 0.40 ± 0.11, OFF: 0.37 ± 0.11, p < 0.01), and the main effect of environment appearance (VR vs. AR) on Landing Position (VR: 0.42 ± 0.08, AR: 0.35 ± 0.13, p < 0.01). No significant interaction was found. The two-way ANOVA indicated no significant differences in any SSQ subscale or Time.

Further, the two-way ANOVA indicated no significant differences in SSQ or Time. Averaged across all four Avoid conditions, the values were SSQ (Nausea: 8.05 ± 11.80, Oculomotor: 15.63 ± 18.46, Disorientation: 20.01 ± 29.23, Total Severity: 16.25 ± 20.37) and Time (2.87 ± 0.54 [sec]).

5.3 Results of StepOn condition

Figure 9 illustrates the outcomes of the author-developed questionnaire, which assessed the perceived ease of walking, fear of walking, and ease of avoidance. A two-way ANOVA revealed a significant main effect of LCDs (ON vs. OFF) on Ease (ON: 4.94 ± 1.61, OFF: 3.72 ± 1.61, p < 0.05) and Step (ON: 4.72 ± 1.85, OFF: 2.91 ± 1.67, p < 0.001). No significant main effect of environmental appearance (VR vs. AR) or interaction was found. Figure 11 shows the results of the head pitch angle. Two-way ANOVA revealed a significant main effect of LCDs (ON vs. OFF) on Average (ON: 17.84 ± 12.46, OFF: 31.33 ± 16.76, p < 0.001) and Variance (ON: 21.36 ± 17.71, OFF: 63.95 ± 79.82, p < 0.05). No significant main effect of environmental appearance (VR vs. AR) or interaction was found. Figure 10 depicts the trajectory of the dominant foot when the participant stepped on the virtual object. The upper-right value indicates the number of trials in each condition where the participant did not successfully step onto the virtual object, and their dominant foot did not make contact with the hit decision.

Foot Distance metric represents the 3D distance between the object and the point of contact where the decision to hit the step object was first made. This data was averaged across four trials for each participant in every condition. However, some participants failed to accurately step onto the object and thus did not make contact with the hit decision. Consequently, we calculated the average after excluding these failed data points. Notably, one participant was unable to step onto the object in any of the four trials under the OffA condition.

Two-way ANOVA did not find any significant differences in SSQ, Foot Distance, or Time. Averaged across all four StepOn conditions, the values were SSQ (Nausea: 9.39 ± 13.86, Oculomotor: 18.00 ± 23.02, Disorientation: 24.79 ± 31.81, Total Severity: 19.23 ± 23.35) and Time (2.85 ± 0.64 [sec]).

6 Discussion

The main findings are as follows:

• H1) The IPQ and ARPQ results did not show a significant effect of the expanded downward FOV on presence, providing no confirmatory support for H1. As in the present experiment, Experiment 1 of Nakano et al. (2021) also reported no significant difference in presence when the downward FOV was increased. The authors attributed this to participants barely moving their bodies and therefore rarely looking down into the downward FOV. In contrast, in Experiment 2 of the same study, where participants performed tasks involving larger body movements, such as touching a door with their hand or crouching down, expanding the downward FOV significantly increased the perceived realism of the virtual environment (IPQ: Real). In the present experiment, participants interacted with the virtual objects by avoiding or stepping on them with their feet, but unlike the previous study, they did not stop and maintain prolonged contact with the objects, and the interaction was momentary and occurred during walking. This difference may explain why expanding the downward FOV did not affect presence in our study.

Alternatively, the effect of the expanded downward FOV may have arisen not from interacting with the objects but from the enhanced three-dimensionality of the virtual objects. In an exploratory analysis, the raw (uncorrected) results showed that the expanded downward FOV improved the IPQ:Real (ON: 3.80 ± 1.04, OFF: 3.25 ± 1.09), ARPQ5 (ON: 5.88 ± 1.15, OFF: 5.31 ± 1.01), and ARPQ7 (ON: 5.50 ± 1.32, OFF: 4.38 ± 1.59) items in the Avoid condition. Both ARPQ5 (“Did you have the impression of seeing the virtual objects as merely flat images or as three-dimensional objects?”) and ARPQ7 (“Did you have to make an effort to recognize the virtual objects as being three-dimensional?”) concern the perceived three-dimensionality of the virtual objects. These differences, however, did not reach significance after the correction across the 22 items.

One possible explanation is that humans estimate the three-dimensionality of objects based on the disparity of visual information obtained from both eyes, and the closer the object is to the body, the more three-dimensional it feels. By presenting images in the downward FOV, objects can be displayed closer to the body, thereby enhancing the three-dimensionality of the virtual objects. This may help explain the presence improvement reported with increased downward FOV in Nakano et al. (2021).

The results of the IPQ and ARPQ in the StepOn condition did not show even an exploratory trend toward an effect of the expanded downward FOV. One possible reason is that the virtual object presented in the StepOn condition (height: 3 [cm]) was lower than the virtual object shown in the Avoid condition (height: 30 [cm]), making it harder to perceive the three-dimensionality. In previous studies (Nakano et al., 2021; Nakano and Narumi, 2025), participants perceived three-dimensionality in the downward FOV through a full-body humanoid avatar overlaid on their own bodies. The experiments were conducted in a more three-dimensional VE than the one used in the StepOn condition in this study, which included virtual objects such as descending stairs, flames, and a mirror. These results suggest that displaying a humanoid avatar or thick virtual objects may influence the perception of three-dimensionality and be effective for image presentation in the downward FOV.

• H2) The results of the author-developed questionnaire indicate that increasing the downward FOV facilitated a better understanding of the situation around the feet, making walking feel easier in the StepOn condition. Participant feedback suggested that the expanded downward FOV made walking easier, as expressed in comments such as “It was easy to walk (OnV)” and “The wider field of view made walking feel easier (OnA).” In the Avoid condition, the effect on perceived ease did not reach significance after correction. Unlike the StepOn task, which required precise foot placement on a specific target, the Avoid task imposed no such constraint, so walking may have felt easy even with a limited downward FOV. The analysis of the fear of walking showed no significant differences; these results do not support H2. One participant nonetheless remarked, “I felt scared stepping over a large object because I could not see within the downward FOV (OffV),” suggesting that the inability to see a large object clearly may contribute to fear. Additionally, the measured average task time in all conditions did not change with the expansion of the downward FOV. Increasing the downward FOV during walking may influence the ease of walking rather than speed.
• H3) The author-developed questionnaire showed that increasing the downward FOV made it easier to avoid or step on the virtual object, with a significant main effect of the downward FOV in both conditions. This was reflected in comments such as “It was easy to recognize my own feet, so it was easy to walk and avoid objects (OnV)” and “The VR feet felt like my own feet, and I was able to step onto the red dot (OnA).” One possible explanation for the task’s perceived ease is that the expanded downward FOV may have made the distance perception to the virtual object more accurate. Several studies have suggested that an enlarged FOV improves egocentric distance perception (Jones et al., 2016), although reviews of the broader literature indicate that the evidence remains mixed (Renner et al., 2013; Feldstein et al., 2020). In this experiment, egocentric distance perception may have been more accurate because the larger downward FOV made it easier to see the ground surface (Flückiger and Baumberger, 1988; Bian et al., 2006; Kluzik et al., 2007), which is important for determining the relative distance of objects. Meanwhile, Masnadi et al. reported that horizontal FOV expansion made egocentric distance perception more accurate, but vertical FOV expansion had no effect (Masnadi et al., 2022). Our future work is to investigate whether the increase in the downward FOV affects egocentric distance perception and the characteristics of the downward FOV within the vertical FOV.

In contrast, the trajectory of the dominant foot and its distance from the virtual object in the Avoid condition indicate that increasing the downward FOV led participants to move their feet more when avoiding the object, which contradicts the expectation of minimal movement in H3. The nearest neighbor distance between the participant’s dominant foot and the virtual object significantly increased in the OnV and OnA conditions with the expanded downward FOV. Additionally, when the downward FOV was increased, the distance between the landing positions of the foot after stepping over the virtual object significantly increased. One possible explanation is that increasing the downward FOV influenced participants’ planning when avoiding obstacles. When people step over an obstacle with their feet, the avoidance action is planned while approaching the obstacle, and changes in visual conditions during the stepping movement are known to have little effect on the movement itself (Patla et al., 2002). Distance perception also depends on integrating information from the ground surface from near to far distances. When the near ground is not visible, distances tend to be underestimated (Wu et al., 2004). Presenting the near ground within the downward FOV may therefore have allowed participants to perceive distances more accurately, which could have helped them avoid the objects more appropriately.

An alternative explanation is that as the downward FOV increased, the three-dimensionality of the objects was enhanced, and participants tried to avoid the virtual object by moving their feet more. This interpretation is also supported by previous studies showing that when people step over and avoid objects, they tend to lift their feet higher during binocular viewing, when objects appear more three-dimensional, than during monocular viewing (Patla et al., 2002; Hayhoe et al., 2009; Zhao and Allison, 2021). Although the difference did not reach significance after correction, the exploratory ARPQ analyses in our study suggest that the expanded downward FOV enhanced the perceived three-dimensionality of the virtual objects. These findings suggest that when the downward FOV was increased, participants may have perceived the virtual objects as more three-dimensional and moved their feet further to avoid them. However, participants could perceive the virtual objects stereoscopically in their frontal LCD by tilting their heads downward, even without downward FOV. The expanded downward FOV may have made the virtual objects easier to see, leading participants to move their feet more widely to avoid stepping on them.

Additionally, in the condition without the expanded downward FOV, the number of failures to step onto the virtual object increased in the StepOn condition, because participants could not visually recognize that they had failed to step onto it, as shown by the upper-right values in Figure 10. Looking at Figure 7, which shows the trajectory of the feet in the Avoid condition, in the OnV condition with the expanded downward FOV, the feet hardly touched the virtual object, whereas in the OffA condition without the expanded downward FOV, foot trajectories penetrated the virtual object. Moreover, the experimenter reported that during the experiment, a participant proceeded straight without avoiding the object, despite being instructed to avoid the object. When asked why this participant moved straight without avoiding the object, they replied, “I did not notice the existence of the virtual object because there was no image in the downward FOV.” This participant was one of the two participants whose data were not included in the experimental data as they were erroneous. These results indicate that increasing the downward FOV enables participants to accurately avoid or step on virtual objects.

• H4) The data on the participants’ head pitch angles (average) measured during the experiment showed that increasing the downward FOV caused the head angle to turn upward, thus supporting H4. These effects are similar to the results of previous studies in which the limitation of the FOV caused the gaze to look downward (Authié et al., 2017; Turano et al., 2001) and head movements increased (Venturino and Wells, 1990; Knight and Baber, 2007; Taneda et al., 2021). Additionally, increasing the downward FOV reduced the variance of the head pitch angle, indicating that participants could walk, avoid, or step onto virtual objects without significant head movements. However, as pointed out in previous studies (Nakano et al., 2021; 2023), the prototype HMD used in this study did not have a gaze-detection function; thus, it was not possible to investigate whether participants were fixating on the image presented in the downward FOV or processing it peripherally. Nakano and Narumi (2025) used an eye-tracking-enabled HMD with an extended downward FOV and reported a significant increase in the percentage of time looking at the downward FOV when participants viewed virtual spheres placed on the ground. This finding suggests that a similar increase in downward FOV viewing time may have occurred in the present experiment when participants viewed a virtual object placed at their feet.
• H5) No interactions were found in any of the statistical results in Table 4. In other words, the effect of increasing the downward FOV occurs not only in VR but also in VST-AR. The item that showed a significant difference in the environmental appearance (VR vs. AR) in the two-way ANOVA was the distance to the landing position in the Avoid condition. These results partially support H5: they indicate that, when an effect of increasing the downward FOV is present, it occurs in both VR and VST-AR. However, contrary to our expectations, no significant differences between VST-AR and VR were observed in many measures. The results of the foot trajectory when stepping over the obstacle indicate that, in the VR condition, participants placed their feet farther from the object after avoidance than in the VST-AR condition. Possible reasons for the smaller effect of increasing the downward FOV in VST-AR include the lack of stereoscopic vision in monocular VST-AR, the mismatch between the background and virtual objects displayed in VST-AR, and the latency of the images.

From a different perspective, the environment’s intrinsic nature may influence avoidance behavior. Sanz et al. found that users tend to avoid virtual objects more than real objects in an immersive projection environment (Sanz et al., 2015). Therefore, the movement to avoid objects may have been more significant in VR, which displays a virtual environment, compared to VST-AR, which provides real-world images. We believe that continuing the development of HMDs that reproduce the real environment with a wide FOV will enable a detailed investigation of the differences in user movement and behavior between real and virtual environments.

7 Limitations

The HMD with an extended downward FOV proved effective in testing hypotheses about walking but presented several limitations. In the present experimental setup, the walking range was restricted to 3 [m] due to the limited tracking volume of the HMD’s tracking system. As a result, only unidirectional walking over a short distance could be measured, making it difficult to examine the effects of downward FOV expansion on long-distance walking. The short walking distance also limited participants’ ability to plan obstacle avoidance using visual information from farther away. As a result, the present findings mainly focus on interactions in the peripersonal space, in which users directly observe and step over or onto obstacles. The effects of an increased downward FOV on long-distance walking require further investigation. Furthermore, the types and sizes of the virtual objects were restricted. Therefore, the generalizability of the findings may be limited. Participants completed four practice tasks with an expanded downward FOV prior to the experiment. These tasks were intended to familiarize participants with the downward FOV presentation, and we limited the number of practice sessions to minimize their impact on the experimental tasks. However, since no practice was conducted in the condition without the downward FOV, we cannot rule out that the performance improvement observed in the downward FOV condition was solely due to the prior practice.

The additional optical system increased the HMD’s weight by 200 [g] and may have tilted the head downward. The added housing also increased the distance between the user’s eyes and the eyepiece, reducing the original viewing angle of the VIVE. Additionally, the housing separates the front view from the added LCD, similar to the rim of glasses. The prototype HMD did not have a gaze-detection function. As a result, we cannot confirm when participants utilized the downward FOV during the tasks or whether they processed the displayed objects through central or peripheral vision. These limitations could be mitigated using a dedicated fused lens, reducing weight, and modifying the tracking system.

The fact that most of the experiment’s participants were young people and that the dominant foot of most participants was the right foot may have influenced the experimental results. As discussed for H1, the improvements in realism and the perceived three-dimensionality of the virtual objects did not reach significance after correction for multiple comparisons; these observations are therefore exploratory and require confirmation in future studies. Although a virtual shoe avatar was used in this experiment, a human avatar might make it easier to compare the virtual objects with one’s own body, which could enhance realism and the perceived three-dimensionality. At the same time, a human avatar would reduce the visibility of the feet and their surroundings, which might weaken these effects. The discussion of H2 relies on the author-developed questions (Ease, Fear, and Avoid/Step), which were single-item measures and did not undergo psychometric validation. The sample size (n = 16) may limit the detection of smaller effects, such as the interaction between App (environment) and LCD (downward FOV) examined in the discussion of H5. Although the questionnaire-based findings remain exploratory due to these limitations, the consistent results for head orientation and foot movement indicate that increasing the downward FOV meaningfully affects walking and interacting with virtual objects.

8 Conclusion

In this study, we explored the impact of increasing the downward FOV using an HMD on various factors such as presence, walking behavior, virtual object avoidance/stepping behaviors, cybersickness, and head movement patterns, especially focusing on the visibility of the feet, which is challenging to display with a conventional FOV. We also investigated whether these effects were consistent across both VR and VST-AR settings.

The results of our experiments demonstrated that increasing the downward FOV led participants to move their dominant foot farther when avoiding large virtual objects. Exploratory analyses further suggested that the increased downward FOV may have enhanced the realism and the perceived three-dimensionality of the objects, although these effects did not reach significance after correction. Importantly, this expansion did not lead to increased cybersickness, and participants were able to avoid or step on the obstacles without needing to tilt their heads downward. Moreover, participants reported that walking and interacting with the virtual objects felt easier when the downward FOV was expanded, without affecting their walking speed. Our exploratory analyses suggest that the perceived three-dimensionality of the virtual objects may contribute to the observed effects on foot movement. These effects were observed not only in VR but also in AR. Furthermore, participants placed their feet farther from the object after avoidance in the VR condition than in the VST-AR condition.

In future research, we aim to increase the vertical FOV of the VST-AR, which currently has a narrower FOV compared to VR. We also plan to use eye-tracking technology with a wide vertical FOV to gain deeper insights into how individuals utilize a downward FOV.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. ChatGPT (GPT-5.5, OpenAI) and Claude (Claude Opus 4.8, Anthropic) were used to assist in revising the text during the review process. The authors reviewed all content and take full responsibility for the final version of the manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Andrea Helen Mason, University of Wisconsin-Madison, United States

Reviewed by: 舜波 王, Tianjin University, China

Shi Li, Hangzhou Dianzi University, China