Comparing virtual reality and 2D display experiences in structural inspection

Civil structures, such as bridges, buildings, dams, and tunnels, are essential components for our society and require routine inspections due to their long service periods and exposure to harsh environmental conditions. Recent advances have enabled the development of virtual reality (VR)-based inspection methods, which have primarily been evaluated in terms of their technical accuracy and efficiency. This study was designed to expand this literature by examining the psychological and cognitive experiences associated with VR in structural inspection from a social science perspective. To achieve this, 107 college students were randomly assigned to experience a bridge inspection module either through an interactive VR environment or a non-immersive 2D display presentation of the same virtual inspection environment. We predicted that participants in the VR condition would report greater immersion, higher cognitive loads, greater positive affect, increased heart rate, and better recall of structural geometry compared to those in the 2D display condition. Results generally supported these predictions, except for geometry accuracy measures. Compared with the 2D display condition, the VR condition produced greater positive affect, immersion/presence, Germane cognitive load, and feelings of safety. Heart rate was marginally significant. Perceptions of negative affect, stress, frustration, difficulty, and accuracy of the structure/environment were not different between conditions. These findings provide new human-centered insights into VR-enabled structural inspection and suggest that VR may support future civil engineering inspection, training, and digital twin workflows.

Civil structures such as bridges, buildings, dams, and tunnels are critical for the functionality of our society but may be prone to deterioration over time due to long service periods and exposure to harsh environmental conditions (ASCE, 2025). Therefore, developing effective inspection techniques to ensure their structural integrity is essential. Traditionally, civil structure owners hire professional technicians to perform field visual inspections to record and monitor structural damage over time (Fu, 2005). While being necessary, these practices could be costly, labor-intensive, and prone to errors (Agdas et al., 2016). Recent advances in Structural Health Monitoring (SHM) have introduced new technologies that enhance the efficiency and safety of civil structure inspection (Kong and Li, 2019; Zhang et al., 2023; Farrar et al., 2025). For example, Unmanned Aerial Vehicle (UAV)-based inspection methods (Jordan et al., 2018; Greenwood et al., 2019) have become increasingly common due to their ability to rapidly collect visual data from difficult-to-access locations. Furthermore, when coupled with computer vision, photogrammetry, and/or LiDAR technologies, these systems can also generate accurate 3D representations of civil structures for detailed damage assessment (Zhao et al., 2021; Kaartinen et al., 2022). More recently, Virtual Reality (VR) (Attard et al., 2018; Omer et al., 2019; Omer et al., 2021; Luleci et al., 2024; Kong et al., 2025; Yiğit and Uysal, 2025; Garcia et al., 2026) has emerged as a practical extension of these 3D modeling technologies, allowing users to evaluate 3D structure models within immersive virtual environments from a safe office setting. Such VR studies are also conceptually aligned with emerging digital twin applications in civil infrastructure (Yiğit and Uysal, 2025), where engineers increasingly interact with virtual representations of physical structures for monitoring, visualization, and decision-making support (Willcox et al., 2024).

Despite the success of VR in structural inspection, there is limited data examining the psychological effects of this technology on users and how the experience is similar to, or different from, traditional field visual inspections. This study was proposed to extend the existing knowledge base by investigating the psychological and cognitive experiences of VR users interacting with structural inspection environments compared with an equivalent 2D display condition. Specifically, this study examines differences in emotional response, immersion, cognitive load, recall of structural geometry, perceived safety, and visual perception during inspection-like tasks. To isolate the effects of immersion in the virtual world, both VR and 2D display conditions presented identical inspection content following the same viewing sequence; the primary difference between conditions was whether participants experienced the environment immersivity through a VR headset or non-immersivity through a 2D display. The findings of this study may provide the broader civil engineering community with human-centered insights relevant to the future development of VR-enabled structural inspection workflows.

Psychological advantages of VR and 3D visualization technologies from fields relevant to structural inspection

Motivation

In the context of structural inspection in civil and construction engineering, VR has been considered a promising alternative compared to traditional field visual inspections for several reasons. First, VR-based inspections can be conducted in an office environment, which significantly reduces the costs associated with field visits, such as travel and labor expenses (Omer et al., 2019; Omer et al., 2021; Attard et al., 2018). Second, VR models can be established via data collected by automated systems (e.g., UAVs, LiDAR), covering traditionally inaccessible areas of the structures in the inspection process (Omer et al., 2019; Omer et al., 2021). This also reduces the risk associated with accessing hazardous locations during field visits (Bacco et al., 2020).

Nevertheless, the findings above stem from investigations in engineering disciplines and focus largely on technical and operational advantages. It is also crucial to examine the implications of VR in structural inspection from a social science and human-centered perspective, as the ultimate end users for the established VR modules are humans. Although some existing VR studies in structural inspection (Omer et al., 2019; Omer et al., 2021; Bacco et al., 2020) acknowledges the immersive nature of VR environments, the psychological mechanisms through which immersion and intuitiveness may provide advantages over traditional field visual inspection workflows have not been thoroughly explored yet. This historically overlooked gap highlights the need for an interdisciplinary approach to better understand the human experience associated with VR-based technologies in structural inspection.

Since there is no literature directly investigating the psychological effects of VR in structural inspection to the best knowledge of the authors, the remainder of this section reviews studies from outside traditional engineering disciplines that examine the psychological benefits of VR and 3D visualization technology associated with inspection-like tasks. The findings from these studies may provide transferable insights for VR-based structural inspection. Through a literature review, three key benefits of VR relevant to structural inspection will be discussed: 1) creating immersive 3D environments; 2) reducing psychological stress; and 3) enhancing model visualization. Figure 1 intuitively illustrates the existing and new findings of VR relevant to structural inspection.

Creating immersive 3D environments

3D structural models are commonly viewed on 2D media displays such as computer monitors, tablet screens, televisions, or smartphone screens. Despite their great accessibility, a critical concern is that these 3D models may appear flat on 2D media displays, lacking enriched depth information. This limitation may negatively affect the user’s ability to uncover structural damage. Therefore, it is critical to understand the benefits and effectiveness of inspecting a structure model in an immersive 3D virtual environment compared with a 2D media display.

There have been many VR and 3D visualization technology studies outside of engineering fields that have compared 2D versus 3D experiences and show the superiority of 3D approaches (Renganayagalu et al., 2021; Kuhne et al., 2023). In laparoscopic surgery, surgeons completed a surgical comfort scale and reported that using 3D imaging powered by a 3D flexible camera promoted more positive perceptions of safety and task efficiency compared to using 2D imaging (Buia et al., 2017). In another study exploring the surgical implications of 3D vs. 2D endoscopic ear surgery, surgeons reported a preference for the 3D technology using passive polarization glasses due to depth perception advantages and improved views compared to the 2D technology (Molinari et al., 2020). Enhanced feelings of safety and task efficiency, as well as improved depth perception, are some of the advantages of using 3D models over 2D displays.

Reducing psychological stress

VR-based methods also offer several psychological advantages over traditional methods of inspection, including potential reduction in psychological stress and cognitive mistakes, as well as improved focus, attention, visualization, and feeling of confidence in the VR assessment environment. In a study led by Slater (2009), participants have been shown to exhibit realistic behavior in immersive virtual environments when they experience the sensation of being in a real place (place illusion) and the situation is actually occurring (plausibility illusion), although they are also aware they are not really there.

Psychological stress may be experienced when working in a potentially hazardous environment in the field, such as inspecting tall bridges, being in confined spaces, or being exposed to adverse weather conditions. When individuals are stressed, they often make mistakes or make quick assessments to escape a perilous situation without considering all options (Keinan, 1987). Similarly, focus and attention can be directed specifically to the inspection task if distractions from external factors such as noise, wind, movement, scents, and lighting can be controlled in VR simulations. It takes considerable cognitive resources to face the environmental challenges, which may lead inspectors to make mistakes or miss details in the field. While VR environments are evolving to be more realistic and immersive, users may still feel safe knowing the physical danger from the environment is not a concern.

Confidence is another important aspect of performance in any job and can be increased by training in interactive, immersive environments before conducting real-world inspections. Researchers developed an interactive VR program for embedded electric circuits training and measured how well participants could apply the knowledge to real-world hands-on applications (Wang et al., 2023). Learners reported lower anxiety and higher confidence levels in the hands-on practical task. This study shows VR can increase confidence and improve the competence of inspectors when faced with complex inspections.

Enhancing model visualization

VR and 3D visualization technologies also offer enhanced visualization. Human vision is limited, but with immersive views in a VR environment, inspectors can see multiple angles, zoom in or out on details, view cross-sections easily, and manipulate the model of the structure, thus enhancing the overall visual experience. In a VR training experiment for a visual scanning task, researchers found that visual realism for tasks that involve visual scanning or visual search in visually complex environments is essential (Ragan et al., 2015). These findings provide promising insights into the potential benefits of VR-based structural inspection.

For instance, in a 2D media display configuration for inspecting a 3D structural model, users must face a difficult situation to balance the tradeoff between viewing the structural surface in detail by zooming in the 3D model on screen, compared with obtaining a larger field of view to gain overall spatial awareness by zooming out the model. The need to frequently zoom in or out can disrupt the inspection process. On the other hand, immersive VR allows users to seamlessly navigate and inspect structures without such interruptions.

In a different study performed by Hidding et al. (2024), medical students with no prior laparoscopy training were placed into different groups with 3D and 2D visualization with conventional or complex curved instruments. When using complex curved instruments, students performed suture and a transfer task better in the 3D training group using a stereoscopic 3D display system compared to the 2D group. In exercises where depth perception is necessary, the 3D visualization provided greater advantages. Lastly, in forensic examinations of the injury of a human being, photogrammetric measurements in 3D VR models were found to be more accurate than the estimates made with standard 2D forensic photographs, as reported by Koller et al. (2019). The 3D models in VR environments provided unique visualizations for greater knowledge and perception of injuries of human beings, even after injuries had healed. The findings of this study are encouraging, as the visualization advantages offered by 3D photogrammetry modeling may be especially useful for structural inspections where enhanced visual detail and depth perception are essential for visually finding damage and deterioration.

Hypotheses

Based on past research findings, we predicted that VR users who interacted with an inspection module of a structure would be able to recall structure geometry information more accurately, feel more immersed, and report a higher cognitive load than those who interacted with the structure model in a 2D display. We also wanted to compare the emotional reactions of participants in VR and 2D display settings. We predicted VR users would report greater positive affect and increased heart rate when exposed to a VR environment compared to the same environment in a 2D display. Additionally, we explored negative affect, individual emotions, stress, perceived difficulty, safety, and frustration ratings between the conditions.

Participants

One-hundred and fifteen college students were recruited from General Psychology courses at Coastal Carolina University in Conway, South Carolina, United States of America, using SONA Systems, an experimental participant management system, and offered one research credit point for their participation (4 research credits were required, or an alternative activity). Participants with partial or incomplete data were removed from the sample (n = 8), resulting in a final sample of 107 (see Figure 2). The sample was predominantly women (81.3%), and the average age was 18.53 (SD =.87). Considering race/ethnicity, the majority of the sample identified as white/Caucasian (75.7%), 8.4% identified as Black/African American, 11.2% identified as Hispanic/Latino/Latina, 1.9% Alaskan Native or Native American, and 2.8% Asian. A variety of majors, including science, education, social sciences, humanities, and business, were represented in the sample. The majority of participants were first year (82.2%), 13.1% were sophomores, 1.9% juniors, and 2.8% seniors. The research was reviewed and approved by the college’s institutional review board (IRB). American Psychological Association (APA, 2002) ethical guidelines were followed throughout the study.

Materials and procedure

Participants came to the lab space, signed an informed consent form, and completed a brief tutorial on how to use the VR headset and navigate the VR module, or what to focus on in the 2D video. Participants were randomly assigned to either the VR headset (n = 60) or the 2D display (n = 47) condition, as shown in Figure 2. After the trial, participants from both groups completed paper surveys about their current mood, stress/frustration, immersion/presence, cognitive load, recall of the structural shape and environment, the general VR experience, and demographics.

VR and 2D display conditions

The research team established the VR and the corresponding 2D display conditions used in this study. Figure 3 shows the workflow of the VR module development process. Taleyfac Spanish Bridge, a historic stone-arch bridge in Guam, was selected as the target structure. The workflow starts with UAV-based image collection during a field visit. A large volume of images was acquired, and three representative examples are shown in Figure 3a. Next, a photogrammetry software package (Agisoft Metashape Professional, version 1.6.2, Agisoft LLC, 2020) was used for reconstructing a 3D point cloud of the bridge (Figure 3b). The small blue patches shown in the figure represent UAV camera positions used during the reconstruction process. Based on the established point cloud, a textured mesh model was rendered in Agisoft Metashape and subsequently imported into a computer graphics engine (Blender, version 3.5, Blender Foundation, 2023). As shown in Figure 3c, a virtual 360-degree panoramic camera was positioned at five different viewpoints in the Blender model, including one start/end location in front of the bridge providing an overall view, and four additional locations (labeled 1 to 4 in the figure) providing close-up views of the bridge surface for detailed inspection.

Next, a 360-degree panoramic image was rendered in Blender for each viewpoint, resulting in five panoramic scenes in total (Figure 3d). These panoramic scenes were then imported into an online VR editing platform (VIAR Inc, 2025), where interactive transitional hotspots were added to connect the scenes and establish the virtual inspection workflow (Figure 3e). As shown in the figure, users begin the VR experience at the Starting Scene, navigate sequentially through Locations 1 to 4 to inspect different portions of the bridge, and finally return to the Ending Scene to complete the virtual inspection journey. Further details regarding the VR module development are provided in Kong et al. (2025).

Compared with the VR module presented in Kong et al. (2025), the scene transition design in this study was modified to follow a fixed linear navigation sequence. Rather than freely moving between viewpoints in a self-paced manner, users were guided through the VR environment in a predefined order from the Starting Scene to Locations 1 to 4 and finally to the Ending Scene (see Figure 3e). This design was intentionally adopted to ensure that participants in the VR condition experienced a viewing sequence comparable to that of participants in the 2D display condition, who watched a prerecorded walkthrough video on a television screen. Lastly, the completed VR module (Figure 3f) was loaded onto a Meta Quest headset for participants assigned to the VR condition.

For the 2D display condition, a team member first navigated through the completed VR module explained above, following the same predefined linear sequence used in the VR condition (see Figure 3e). The Meta Quest headset’s built-in screen recording feature was used to capture the entire virtual walkthrough, resulting in a 2-min and 44-s video clip. This design was intentionally adopted so that both conditions presented identical structural content while differing primarily in the level of immersion experienced by participants. This recorded walkthrough video was subsequently displayed on a large-screen television (55-inch Liquid Crystal Display) for participants assigned to the 2D display condition. Participants in this group were told, “In this scenario, you are an inspector performing a structural inspection of a stone arch bridge. A ground team has collected drone images for establishing a 3D bridge model, based on which a walkthrough video of the bridge was created for you. Your task is to view the video to gain a 3D perception of the bridge.”

Figure 4 shows representative frames and navigation timeline from the prerecorded walkthrough video used in the 2D display condition. During the recording process, the team member spent approximately 13–39 s on each of the six scenes, with an average of 34 s spent across the four close-up scenes (Locations 1 to 4). Representative frames from each scene are shown in Figures 4a–f. The visual content presented in the 2D display condition was intentionally designed to be identical to that of the VR condition. However, participants in the VR condition actively navigated between scenes by selecting interactive hotspots using handheld controllers, whereas participants in the 2D condition passively viewed the prerecorded walkthrough video on the television screen. Due to space limitations, screenshots from the VR condition are not separately presented, as their visual context was nearly identical to the representative frames shown in Figures 3a–f.

Fitness tracker

A fitness tracker was used to measure heart rate at the beginning and end of the experiment for both VR and 2D display conditions. We used a basic fitness tracker purchased through Amazon (Jawploy Fitness Tracker with Heart Rate Monitor/Step Counter/Calories, Activity Tracker Smart Watch Sleep Monitor, IP68), linked to the researcher’s cell phone to capture heart rate. Heart rate measurement was concealed from participants; they could not see their heart rate on the tracker, or on the researcher’s phone.

PANAS

The Positive and Negative Affect Scale (PANAS; Watson et al., 1988) is comprised of 20 adjectives to measure positive affect (PA) and negative affect (NA) using a 5-point Likert scale, ranging from 1 = Very slightly or not at all to 5 = Extremely. Ten items are related to PA, and ten items are related to NA. Participants were asked how much each of the terms described how they felt right now, at that time. The scale has high internal consistency and is a very common measure for mood.

Stress/frustration

Three individual questions were asked about stress and frustration, based on questions used in Kolb and Aiello’s (1996) study on the effects of electronic performance monitoring on stress. Questions were presented with a 7-point Likert-style response scale where higher values indicate greater reports of stress or frustration. The questions included “What degree of stress did you experience while working on the tasks during the study?”, using 1 = no stress to 7 = a great deal of stress; “How would you describe the climate/atmosphere for working on tasks in this study?”, using 1 = not frustrating to 7 = frustrating; and “When you were working on the tasks during the study, how did you feel?”, using 1 = stressed to 7 = not stressed (reverse scored).

Immersion/presence

To measure how immersed or present participants felt experiencing VR and 2D display conditions, we used four questions from Carlin et al.’s (1997) study on treating spider phobia with VR. Participants were asked to rate each of the items on a 7-point Likert-type scale. The questions included “How real did the virtual world seem to you?”, using 1 = about as real as an imagined world to 7 = indistinguishable from the real world, “To what extent were there times when you felt that the virtual world became ‘reality’ for you, and you almost forgot about the real world outside?”, using 1 = at no time to 7 = almost all the time, “Did the virtual world seem more like something you saw, or some place you visited?”, using 1 = something I saw to 7 = some place I visited, and “In the virtual world, I felt more like...”, using 1 = I was standing in the laboratory wearing a helmet to 7 = I was in the virtual room. A higher score of the composite of these four items would indicate greater immersion/presence. The terminology of ‘virtual world’ was applicable to both conditions since they were presented with a virtual world to inspect. Participants from the 2D display condition did not report any confusion about the meaning of the questions related to their assigned condition experience.

Cognitive load

The Cognitive Load Questionnaire (CLQ; Leppink et al., 2014) is a 13-item survey that measures cognitive load using a 10-point Likert-style response scale where 0 = not at all the case to 10 = completely the case. The CLQ has three subscales: Intrinsic, Extraneous, and Germane. Intrinsic cognitive load measures the difficulty/complexity of the task. Extraneous cognitive load measures how the instructional design/distractions may interfere with the task. Germane cognitive load measures the effort invested in the process of learning, such as forming connections between new and prior knowledge. Example items included “The content of this activity was very complex” and “This activity really enhanced my knowledge and understanding of how to deal with the problems covered.” Higher scores indicate greater cognitive load.

Recall of structure and environment

Four additional questions asking about the appearance of the bridge and environment were included to assess recall and understanding of the environment. The questions asked, “Which option best describes the overall appearance of the bridge’s exterior (the part you see)?” (see Figure 5a), followed by three drawing options, “Which option best describes the layout of the bridge deck when viewed from above?” (see Figure 5b), followed by four drawing options, and “Regarding green plants, which statement is correct to the best of your knowledge?” (see Figure 5c), followed by four written descriptions. The first three questions were scored as correct or incorrect. A fourth question asked “Based on your experience in the virtual environment (or 2D video), how well could you perceive variations in the roughness of the bridge surface? Consider whether the surface appeared completely smooth, slightly textured, or highly textured”, using a 7-point Likert-style response scale using 1 = not at all, 2 = slightly, 3 = moderately, 4 = neutral, 5 = quite well, 6 = very well, and 7 = extremely well.

VR experiences

Three additional questions were asked about the VR experience during this experiment. The first was about past use of VR technology, asking “Have you ever used a VR headset before?” Two additional questions asked about the difficulty and safety of the virtual environment and were rated using a 5-point Likert-type scale where 1 = strongly disagree to 5 = strongly agree. The questions were “It was difficult to navigate this environment?” and “I felt safe in this environment.”

Demographics

Questions about participants’ age, sex, race, major, and year in school were asked after the other measures, at the end of the study.

After completion of the surveys, participants were debriefed and thanked for their participation.

The majority of participants had used a VR headset previously (78.5%). The difference between the ending heart rate and beginning heart rate for each participant was calculated to create a heart rate difference score. Scores for PA (α =.88) and NA (α =.71) from the PANAS (Watson et al., 1988), immersion/presence scores from the four items presented (α =.47), and Intrinsic (α =.83), Extraneous (α =.81), and Germane (α =.93) subscale values from the Cognitive Load Questionnaire (CLQ; Leppink et al., 2014) were calculated. Individual values from the VR experience questions about safety, difficulty, stress, frustration, and roughness of the bridge surface were used as dependent variables. A series of two-tailed independent samples t -tests were performed to compare these scores between the VR condition and the 2D display condition. Data were assumed to be normal and Levene’s test for group variance was used to determine the equality of variance. Levene’s test indicated that the assumption of homogeneity of variance was violated for three dependent variables: total time, difficulty navigating, and safety. Therefore, degrees of freedom were adjusted using Welch’s correction for these three dependent variables. All other dependent variables met Levene’s assumption of equal variance. The Chi-square Test of Independence was used to analyze the three recall of structure and environment questions that were scored correct or incorrect to determine if condition and response were related for these categorical variables.

Total time

The total time (seconds) spent watching the VR module was similar between the VR (M = 172.32, SD = 79.72) and 2D display conditions (M = 161.55, SD = 10.49), t (61.60) = 1.04, p =.31, d = −0.18, 95% CI [-.56,.20]. There was greater variability in time spent in the VR session, which is expected since participants had to actively click to move between areas.

Heart rate change

Heart rate difference scores indicated that participants showed an increase in heart rate from the beginning to the end of the experiment, after viewing the VR module or watching the 2D video. Heart rate increase was greater for the VR condition (M = 6.62, SD = 15.06) compared to the 2D display condition (M = 1.19, SD = 16.01), t (105) = 1.80, p =.08, d = −0.35, 95% CI [-.73,.04]. However, this difference was only marginally significant.

PANAS

Participants who experienced the VR module in the Meta Quest (M = 2.73, SD =.71) reported greater PA compared to those who watched the prerecorded 2D video (M = 2.08, SD =.68), t (105) = 4.81, p <.001, d = −0.94, 95% CI [-.1.34, −0.53]. Participants in the VR condition (M = 1.30, SD =.38) reported similar NA compared to those in the 2D display condition (M = 1.31, SD =.30), t (105) =.09, p =.93, d =.02, 95% CI [-.36,.40]. See Figure 6 for results. The individual terms from the PANAS were also compared between both conditions as an exploratory analysis, and their results are presented in Table 1. Each of the PA terms was rated significantly higher in the VR condition compared to the 2D display condition, except strong and alert which were only marginally significant. None of the NA terms were significantly different between conditions.

Stress/frustration

For the question “What degree of stress did you experience while working on the tasks during the study?”, those in the VR condition (M = 1.75, SD = 1.19) reported similar low scores compared to those in the 2D display condition (M = 1.79, SD =.95), t (105) =.18, p =.86, d =.03, 95% CI [-.35,.42]. For the question “How would you describe the climate/atmosphere for working on tasks in this study?”, participants in the VR condition (M = 1.70, SD = 1.09) reported similar low ratings of frustration compared to those in the 2D display condition (M = 1.66, SD = 1.20), t (105) =.18, p =.86, d = −0.04, 95% CI [-.42,.35]. When answering “When you were working on the tasks during the study, how did you feel?”, respondents in the VR condition (M = 2.20, SD = 1.68) reported similar low levels of stress compared to those in the 2D video condition (M = 2.36, SD = 1.82), t (105) =.48, p =.64, d =.09, 95% CI [-.29,.48].

Immersion/presence

For the composite score of the four immersion/presence items, participants in the VR condition (M = 3.35, SD = 1.11) reported higher scores compared to those in the 2D display condition (M = 2.73, SD = 1.00), t (105) = 3.03, p =.003, d = −0.59, 95% CI [-.98, −0.20].

Given the low internal consistency of the immersion/presence scale in our sample, we calculated Cronbach’s alpha separately by group and found they were also similar (VR group α =.46, 2D display group α =.40). For additional information, we investigated each question separately. Levene’s test indicated that the assumption of homogeneity of variance was violated for question two and degrees of freedom were adjusted using Welch’s correction for this dependent variable. For the question “How real did the virtual world seem to you?”, respondents in the VR condition (M = 3.37, SD = 1.55) reported similar scores to those in the 2D video condition (M = 3.40, SD = 1.84), t (105) =.12, p =.91, d =.02, 95% CI [-.36,.40]. For the question “To what extent were there times when you felt that the virtual world became ‘reality’ for you, and you almost forgot about the real world outside?”, respondents in the VR condition (M = 2.55, SD = 1.58) reported higher scores those in the 2D video condition (M = 1.45, SD =.93), t (98.09) = 4.51, p <.001, d = −0.83, 95% CI [-1.22, −0.43]. For the question “Did the virtual world seem more like something you saw, or some place you visited?”, respondents in the VR condition (M = 3.18, SD = 2.01) reported higher scores those in the 2D video condition (M = 2.49, SD = 1.60), t (105) = 1.93, p =.06, d = −0.38, 95% CI [-.76,.01]. For the question “In the virtual world, I felt more like...” respondents in the VR condition (M = 4.32, SD = 1.98) reported higher scores those in the 2D video condition (M = 3.57, SD = 2.10), t (105) = 1.87, p =.06, d = −0.37, 95% CI [-.75,.02].

Cognitive load

The CLQ (Leppink et al., 2014) had three subscales: Intrinsic, Extraneous, and Germane. For Intrinsic cognitive load, those in the VR condition (M = 1.66, SD = 1.77) reported similar scores compared to those in the 2D display condition (M = 1.51, SD = 2.07), t (105) =.42, p =.67, d = −0.08, 95% CI [-.46,.30]. For Extraneous cognitive load, respondents in the VR condition (M = 1.06, SD = 1.67) reported similar scores compared to those in the 2D display condition (M = 1.37, SD = 1.97), t (105) =.89, p =.38, d =.17, 95% CI [-.21,.56]. For Germane cognitive load, participants in the VR condition (M = 3.57, SD = 2.63) scored significantly higher than those in the 2D display condition (M = 2.17, SD = 2.55), t (105) = 2.78, p =.006, d = −0.54, 95% CI [-.93, −0.15]. See Figure 7 for results.

Recall of structure and environment

For the question “Which option best describes the overall appearance of the bridge’s exterior (the part you see)?”, 28.33% answered the question correctly in the VR condition, and 29.79% answered it correctly in the 2D display condition. The relationship between condition and score for this question was not statistically significant, χ 2 (1, N = 107) =.03, p =.87, φ =.02. For the question “Which option best describes the layout of the bridge deck when viewed from above?”, 30.0% answered the question correctly in the VR condition, and 34.0% answered it correctly in the 2D display condition. The relationship between condition and score for this question was not statistically significant, χ 2 (1, N = 107) =.20, p =.66, φ =.04. For the question “Regarding green plants, which statement is correct to the best of your knowledge?”, 21.67% answered the question correctly in the VR condition, and 19.15% answered it correctly in the 2D display condition. The relationship between condition and score for this question was not statistically significant, χ 2 (1, N = 107) =.10, p =.75, φ =.03.

For the question “Based on your experience in the virtual environment (or 2D video), how well could you perceive variations in the roughness of the bridge surface? Consider whether the surface appeared completely smooth, slightly textured, or highly textured”, respondents in the VR condition (M = 4.35, SD = 1.29) reported similar scores compared to those in the 2D display condition (M = 4.17, SD = 1.46), t (105) =.68, p =.50, d = −0.13, 95% CI [-.51,.25].

VR experiences

For the question “It was difficult to navigate this environment?” participants in the VR headset condition (M = 1.92, SD =.72) reported significantly lower scores compared to those in the 2D video condition (M = 2.53, SD = 1.12), t (74.51) = 3.27, p =.002, d =.67, 95% CI [.28, 1.06]. For the question, “I felt safe in this environment”, participants in the VR condition (M = 4.10, SD =.73) reported significantly higher scores compared to those in the 2D display condition (M = 3.74, SD =.85), t (91.09) = 2.29, p = 0.02, d = −0.45, 95% CI [-.84, −0.07].

This study investigated differences in emotional response, immersion, cognitive load, perceived safety, and visual perception between VR and 2D display conditions when users perform tasks under the same virtual structural inspection environment of a bridge structure. We predicted that VR users would be able to identify structure geometry more accurately, feel more immersed, and report a higher cognitive load than those who were in the 2D display condition. We also predicted that users would report greater positive affect and increased heart rate when exposed to a VR environment compared to those who were in the 2D display group. We found support for our predictions, excluding measures of geometry accuracy and heart rate.

Heart rate increased in both conditions, and the increase was greater in the VR condition compared to the 2D display condition. However, this difference was only marginally significant. Participants who experienced the VR module also reported greater PA, while negative NA, such as distressed, upset, and afraid ratings, remained similar between groups (see Table 1). The effect size for PA was large (d =.94), suggesting the groups do not overlap and the finding has a strong practical impact. Looking at the individual terms that made up PA, participants reported feelings more alert, active, attentive, interested, and excited when viewing the module using the VR headset. Many of the individual PA terms also had large effect sizes. Although increased heart rate alone may reflect general arousal rather than excitement specifically (Draghici and Taylor, 2016; Baevsky and Chernikova, 2017), the combination of elevated PA scores and unchanged NA scores suggests that the heightened physiological response observed in the VR condition was more closely associated with positive engagement and excitement. From a practical perspective, these findings suggest that VR-based inspection environments may promote greater user engagement and attentional involvement during inspection-like tasks.

Participants in the VR condition also reported significantly higher immersion/presence ratings than those in the 2D display condition, as predicted, and the associated effect size for immersion/presence was medium. However, the composite measure we used yielded low internal consistency, so these results should be interpreted cautiously. Individual item analysis revealed only the second question yielded statically significant differences between groups, and questions three and four showed only marginally significant differences. The response wording for question one, which was not significantly different between groups, may have been unclear. The low internal consistency of this presence measure in our sample was reported by both conditions, suggesting that the 2D display group did not judge it to be irrelevant or confusing. The low internal consistency may be due to an artifact of how the scale was made, with a small number of only four items. Future studies may investigate alternative presence/immersion scales that are longer and more robust for comparison.

Cognitive load for the Intrinsic and Extraneous subscales from CLQ (Leppink et al., 2014) was similar between conditions, but Germane cognitive load was significantly higher in the VR condition than the 2D display condition, with a medium effect size. Intrinsic cognitive load is primarily about the inherent complexity of the task, while Extraneous cognitive load is associated with unclear instructions, distractions, or other factors that make the task hard to interpret. In contrast, Germane cognitive load reflects productive mental effort that supports learning, understanding, reflection, and application of concepts (Leppink et al., 2014). The higher Germane cognitive load observed in the VR condition suggests that participants were more cognitively engaged with the inspection activity without perceiving the task itself as more difficult or confusing.

Accuracy questions about the structure geometry (see Figures 5a,b) showed there was not a relationship between condition and correct response, with about a quarter of the sample answering the questions correctly overall. The relatively low recall accuracy observed in both groups may be explained by several factors. First, participants were not instructed beforehand to memorize the shape of the structure or informed that geometry-related questions would be asked questions at the conclusion of the test. Second, the participants are undergraduates from a variety of majors who may not have sufficient knowledge to interpret bridge drawings in plan and elevation views. Similarly, for the environmental recall question (see Figure 5c), the VR condition showed slightly higher recall accuracy than the 2D display condition, but the relationship between condition and correct response was not significant. Participants in the VR and 2D display conditions also reported similar perceptions of bridge surface roughness, suggesting that the immersive advantages of VR did not substantially improve perception of texture-related visual details in the current VR module. The VR and 2D display conditions were intentionally designed to present nearly identical visual content, viewing sequences, and viewing durations, which may also have reduced the likelihood of large differences in texture perception between conditions. Lastly, participants in the VR and 2D display conditions reported similar responses to the stress and frustration questions. In the exploratory analyses, difficulty in navigating the environment was rated significantly higher in the 2D display condition compared to the VR condition, but those in the VR condition reported feeling safer than those in the 2D display condition.

Collectively, these findings suggest that VR-based inspection environments may provide meaningful human-centered advantages over non-immersive presentations of the same virtual inspection content. Although the VR condition did not significantly improve the accuracy of structural shape and environmental recalls in the current study, participants reported greater immersion, positive affect, Germane cognitive load, and perceived safety without increased stress, frustration, or navigation difficulty. If conservative Bonferroni corrections for multiple comparisons were applied, Germane cognitive load and perception of safety were the least resilient findings, whereas PA, immersion, and navigation difficulty proved most robust. These experiential advantages reported by this study may support training effectiveness, sustained attentional involvement, and user acceptance of VR-enabled inspection systems. Such human-centered interaction may also fit naturally within emerging digital twin frameworks in civil infrastructure, where engineers remain actively involved in interpreting virtual representations of physical structures and providing feedback to support real-world inspection and decision-making processes.

Limitations and future research

Although the findings of this study are encouraging, several limitations should be acknowledged. First, our sample was comprised of college students, the majority of whom were women and Caucasian, which may limit the generalizability of the findings. Future studies should investigate more diverse populations and explore potential demographic differences in VR inspection experiences. Nevertheless, the structural module used in this study required minimal civil engineering background knowledge because participants were asked to make general visual observations rather than conduct a formal structural condition assessment.

Second, we were limited by the VR module in this study, which was a historic stone arch bridge. Other structures, such as buildings, dams, and tunnels, may produce similar or different outcomes. Third, the comparison was limited to an immersive VR condition and a non-immersive 2D presentation of the same virtual inspection environment. Although this design controlled the visual content and viewing sequence between conditions, it did not fully isolate interactivity, physical head movement, and potential novelty associated with VR users. Consequently, the observed differences between conditions may reflect the combined influence of these factors rather than immersive display alone. Fourth, the 2D display condition may not fully represent current civil engineering inspection workflows, in which inspectors typically interact with raw field images or navigable 3D models on computer monitors, or conduct inspections in the field. Future work could incorporate a real-world field inspection as a comparison condition. Measuring affect, accuracy, immersion, stress, cognitive load, and other variables when inspectors complete their inspections in the real world and comparing these variables to the VR condition would be an interesting progression of this research.

Lastly, the physiological measures used in this study could be expanded in future studies. More precise instruments, such as skin conductance, EKG, pupil dilation, and other body reaction measurements, may provide additional insight into user responses during VR inspection. Future studies could also include additional measures related to confidence, focus, attention, and inspection performance.

Compared to a 2D display condition, experiencing a civil structure inspection module through a VR headset produces greater positive affect, immersion/presence, Germane cognitive load, and feeling of safety. While heart rate was greater in the VR headset condition, the difference was only marginally significant. In contrast, perceptions of negative affect, stress, frustration, difficulty, and accuracy of the structure/environment were not different between the VR and 2D display conditions. These findings can be used by the civil and construction engineering stakeholders to understand the psychological experience in a VR-enabled structural inspection environment. More broadly, the results suggest that VR-based inspection systems may support engagement, training, and user interaction within future civil engineering inspection and digital twin workflows.