Avatars are often used in video games to serve as fun alter egos and show off custom designs, but scientific research could also now back their use. A new study from University College London shows the use of avatars in virtual reality (VR) heightens a sense of virtual presence, or of feeling physically and naturally situated within one’s digital environment. The findings, though still in their early stages, have strong implications for how developers should be designing VR applications, especially when it comes to training programs.

Notably, this is one of the first VR research experiments done entirely “in the wild,” or in authentic, uncontrolled settings away from the lab.

Presence is important in VR because it gives users subtle clues on how to interact with the surrounding computer-generated world, which impacts how convincing it is. It can act as a bridge between a user’s real behaviors and expectations, and VR’s fictional settings.

Avatars—or virtual bodies physically embodying the user—are one of the most widely inconsistent design aspects in VR, with some developers adopting them and others not. Without research to better understand when and how to use an avatar to affect presence, the discrepancy risks disorienting and confusing users across the VR gamut.

“Getting users to behave naturally is more important than visuals alone in making the VR experience feel real,” said Anthony Steed, lead researcher. “If the building blocks of presence aren’t right, it can derail the illusion.”

Specifically, the research team tested three hypotheses on virtual presence, including and going beyond avatars:If the participant in the experiment had an avatar, this would have a positive impact on presence

  1. If the participant in the experiment had an avatar, this would have a positive impact on presence
  2. Having another character in the VR appear to engage in eye contact with the participant would have a positive impact on presence
  3. Having the participant engage in real body movements while in the virtual environment would have a positive impact on presence, as well as body ownership illusion with the avatar

The team tested these by creating a virtual setting in a bar-café where real-world participants could watch a virtual singer perform. Each participant, who could opt-in anonymously via a free app using Samsung Gear VR or Google Cardboard, would be given one of eight design conditions across three pairs: avatar vs. no avatar; singer engaging in eye contact vs. singer not engaging in eye contact; and physical induction, tested by having the user tap his or her hand to the music, vs. no bodily movement. Results were measured by having the participants fill out a questionnaire at the end assessing their feelings of virtual presence.




In the end, the team collected 59 eligible public data sets. After being analyzed, the data confirmed having an avatar did indeed increase self-reported feelings of presence.

“There’s something about seeing yourself that provides a clearer explanation of context, which is underexploited in games and training programs,” said Steed. “For example, you could look down and see the color of a uniform to give a sense of identity, or see your feet planted in water. Companies are experimenting with avatars where you can actually be ‘yourself’, and we’d like to explore this in future work.”

Results were a bit murkier for the other hypotheses. The resolution of the singer’s face ended up not being very clear, so it was difficult to see where she was looking to determine eye contact, skewing the data. And results were surprising for physical induction: when a user was asked to tap along to the music, it actually decreased the sense of presence, rather than increased it as the team expected. Steed attributes this to the inability to monitor whether people were actually tapping, and the participant’s motion thus not necessarily being in synchronization with the animation.

Overall though, the study proves there’s a link between design elements and feelings of virtual presence, encouraging developers to be more thoughtful in the way they deploy elements such as avatars and physical movement. This is especially relevant for professional training and psychological therapy platforms that depend on authentic behaviors to be effective.

From a broader perspective, the study shows that collecting VR data in the wild—which is missing almost entirely from industry literature—is very achievable, and there’s great opportunity to continue building and fine-tuning community guidelines.

“I hope the research community will start having more discussions on ways to develop best practices for VR, and how other labs can run such studies,” said Steed. “Companies are building the hardware, but it’s equally important for early adopters to curate data showing the impact of the technology, as smartphone companies have been doing for years.”

This study is only the beginning for Steed and his team of researchers. Next, they plan to test their hypotheses in the lab where they can use hand-tracking to monitor induction—and better gauge reactions in-person. Within ethical limits, the team would also like to collect objective data (using health monitors to track people’s heart rate, for example) and look at dozens more variables that could shape presence, leading to more comprehensive results.

Read Part 1 about VR time perception, or find out more about presence in virtual reality in IEEE Xplore.