Does visual perception in virtual reality differ from reality?

 
Method

To eliminate the technical factors that would influence gaze results, the eye-tracking devices chosen in this study for R and VR were developed and constructed by the same company (SensoMotoric Instruments, Germany). The eye-tracking technology and the level of accuracy in both conditions were therefore assumed to be the same. Next, 21 young subjects (age from 19 to 29) were recruited and three tasks were designed for this study. The testing room used for accomplishing the tasks in R was modelled with those noticeable and important features in VR (see Fig. 1a), so that the results from both conditions would be comparable.

In Task 1, a series of random targets were shown on the screen in R and on the head-mounted device (HMD) with a HTC VIVE with built-in eye tracker (HTC, Taiwan) in VR. The participants were asked to stare at the targets appearing in the random corners and then move the gaze point back to the centre of the screen. In Task 2, an object was moving along a pre-defined route on the screen and the participants were instructed to follow the object with their gaze all the way to the end. In Task 3, the participants were asked to stare at the middle target on the screen while the monitor was placed at different distances (1, 2 and 3m) from the subjects. The setup in both conditions is shown in Fig. 1.

The gaze data were collected and prepared according to the guidelines of Holmqvist et al. (2015). Finally, a statistical analysis of non-parametric Friedman test was conducted.

Results & Discussion

The results from Task 1 showed that the accuracy is around 0.5 degree of visual angle in R and VR in directional vision, which revealed no significant difference after the statistical test. In addition, the gaze accuracy was better in the centre of the visual field compared to the corners in both conditions. The results were consistent with other previous studies (Hornof et al., 2002; Holmqvist et al., 2012).

In the results from Task 2, a higher gaze accuracy was observed in R than in VR. A possible explanation could be the lower resolution of the HMD, which made the moving object difficult to observe.

In the results of Task 3, there was no significant difference between R and VR when the monitor was one metre away from the subjects. However, a higher accuracy was also observed in R when the monitor was in farther position compared to VR. The lower resolution of HMD could be the reason leading to these results as addressed above.

Future directions

In general, this preliminary study proved that there was no significant difference in gaze accuracy between R and VR, suggesting the similar level of visual perception in R and VR was achieved from this perspective. However, due to the limited resolution of the HMD, some of the results reflected that gaze behaviour may be worse in VR, which should be further investigated in the future using other HMDs with higher resolution. Moreover, the virtual room in this study provided high enough fidelity with important features of the environment (e.g., materials of furniture, relative position of objects) immersing the subjects in familiar surroundings to perform the tasks, which was also a critical factor for implementing a virtual environment for sport training.

Acknowledgements
The project is supported by the German Research Foundation (DFG) under grant WI 1456/22-1. The authors appreciate the assistance from Mats Naujoks, Martin Luca and Dan Bürger for experiment conduction and data collection.