From the learnings made in Designing inclusive VR training part 1, we set out to begin testing some different approaches in building an inclusive learning experience. From the user tests with people with dyslexia and interviews with The Swedish Dyslexia Association, our main takeaway was to reduce the amount of text used in an VR experience, compared to the typical best practice. But what about other neuro diverse people? Do they also have issues with the design of VR applications?

In this second study, we wanted to get more concrete and answer two fundamental questions in regard to design and interface:

  1. How does the level of realism affect the learning experience in VR?
    Our hypothesis : individuals will not learn as efficiently from a less realistic VR environment. (The Swedish Autism- and Asperger Organization highlighted this could be an issue for neuro diverse people).
  2. What is the effect of different instruction methods in VR?
    Our hypothesis: people will learn differently depending on the instruction type.

The study consisted of 2 tests, and was conducted with 25 participants diagnosed with neuropsychiatric disabilities. Out of the 25 participants, 13 were diagnosed with autism spectrum disorder (ASD), 7 with attention deficit (hyperactivity) disorder (ADHD/ADD), and 5 were diagnosed with both.  

In the first test, two 3D art styles were used for the same VR kitchen environment. One was designed to be realistic looking, while the other was created in a simplified and stylized way.

Image 1. Realistic version of kitchen
Image 2. Stylized version of kitchen

In the second test we focused on comparing instruction methods and feedback mechanisms. Three types of instructions were tested: visual, text and verbal.

Each test was focused on packing a picnic basket with items found in a kitchen environment. This environment was chosen to ensure all participants began with a certain level of familiarity.

In both tests, the participants were required to memorize the order, amount and the location where items were placed in the kitchen.


When asked, almost all the participants preferred the realistic environment over the simplified one. Several participants commented that the realistic environment was more attractive and felt better to be inside. Further, many of the participants thought realism was important for their learning experience.

Contrary to preference and expectations, the results showed that the participants performed better in the simplified stylized environment (see Table 1), regardless of which test they started with.

Number of errors for both stylized and realistic environments and in which order they were tested (left to right)
The mean values of how the participants rated the instructions and the level of difficulty for both stylized and realistic environments and in which order they were tested.


The second test focused on instruction delivery methods: Visual, text and audio. While visual instructions were the most preferred method, not everyone felt that visual instructions were the best for them.

Mean rated values for each instruction method. The sorting order refers to which order the participants did each instruction step. The participants answered to: "How would you rate the instructions? 1=bad, 7=great"
Mean rated values for task difficulty. The sorting order refers to which order the participants did each instruction step. The participant responded to: "How would you rate the tasks level of difficulty? 1=hard, 7=easy"
Results from the questionnaire asked after the test was completed and the VR equipment had been put away

To compare the instruction methods,visual methods were the most preferred and highest rated. Interestingly, this is also the method that resulted in the worst performance.

One benefit of visual instruction is that it is intuitive and has little limitations in design choices when compared with text and voice. However, visual instructions are also something that users are generally not used to. Future research coupled with extensive user experience studies will be needed to investigate how visual instructions should be optimally designed.

Text was rated the lowest preferred instruction method, but had the best performance results. It was also the method that many felt worked best for them. instructions containing numbers and bulleted list turned out to be preferable, as were instructions that were divided into smaller bits and displayed one at the time. It is important to note that relying on text as an instruction method may be too much of a challenge for people with reading disabilities.

Verbal instructions were rated in between visual and text instructions for both level of difficulty and of how well the instructions worked for the task. Several participants thought that verbal instructions would work better than it actually did as this resembles how you would learn to perform a practical task from another person.

Many participants felt there was too much information to memorize when following the verbal instruction. Many wanted to begin the task as soon as possible because they were afraid they would not remember otherwise.

Verbal instructions were given from a test leader. Participants could request any instruction to be repeated if needed. Ideally, verbal instructions should be built into a VR application, however, the downside is that without a test leader, the user is unable to ask open questions to the same extent.

When asked if the participant believed Virtual Reality is a good tool to learn new things, 21 out of the 25 the participants said yes. Furthermore, 18 of the participants thought that Virtual Reality was better than theoretical learning.

Our hypothesis that individuals will not learn as efficiently from a less realistic VR environment, was not correct.


Visual instructions was the preferred method, but textual instructions had the least errors made. The majority of participants said they would want a combination of instruction, rather than just one method. A recommended solution would be to give the users an option to choose what type of instructions they want.

In addition, we noticed that a level of personalization can help facilitate a good learning experience . A few participants thought that some colors and lights were too strong, making them lose focus on the task. This could be avoided if colors and lights could be adjusted by the users themselves. We also experienced an “exhaustion” phenomenon. In Test 1, whether you started with a realistic or stylized environment, the performance was worse for the second test in VR. The same pattern was also seen with the different instruction methods in Test 2. This indicated that the participants became fatigued as the tests went on. For future use of VR in learning situations, tasks should be kept short and be repeatable.

The majority of participants felt that VR is a good tool for learning and 72% believed it to be better than traditional learning methods. This indicates that VR could be a helpful tool for people diagnosed with neuropsychiatric disabilities.

To conclude, when virtual reality is optimized with proper introduction, personalized settings, and intuitive instructions, it can become a powerful tool for anyone who wants to learn new skills.

Jakob Way
March 7, 2023