We release Hafnia Hands ¹ , a set of hand textures for use in virtual reality (VR) research. The textures cover all six skin-tone levels of the Fitzpatrick scale (Fitzpatrick, 1988) in addition to three non-human variants.

Hand representations are an increasingly common way for users to interact in virtual environments and the hand appearance an important part of the user experience and bodily self-consciousness when immersed. Our resource allows researchers to easily setup VR studies with visually realistic hand representations based on a standard skin tone scale.

Virtual hand representations are increasingly enabled by consumer-oriented VR technology through built-in hand tracking support (e.g., the Oculus Quest ² ) as well as add-on products that can add hand tracking support to headsets that do not otherwise support it (e.g., the Leap Motion ³ ). While these systems usually provide hand models that work with their tracking, texturing of these models is limited. For example, the Oculus and Leap Motion SDKs both do not include hand textures.

Unfortunately, creating detailed textures for existing hand models requires the expertise of a texture artist—a role not typically found in research teams. There are textured hands available for purchase, but these usually use different underlying 3d models and hence are not directly compatible with the hand tracking data. While it can be possible to adapt or create a hand skeleton that matches a model to that data, that in itself is a task that commonly requires expertise from a 3d artist. For researchers that want to work with hand representations, but do not have expertise in 3d modeling, rigging, and texturing, there hence is a lack of easy to use options.

We commissioned the set of nine hand texture variants from a 3d artist on Fiverr. Fiverr is a platform for connecting with freelancers, focused on creative tasks such as video editing or 3d modeling. The set (shown in Figure 1) includes six different human skin-tones, as well as three non-human hands: a robot hand, an alien hand, and a skeleton hand. Each variant comes with albedo, ambient occlusion, metallic, smoothness/roughness, and normal maps and hence is compatible with commonly used physically-based shading models (such as the Unity Standard Shader).

The textures are designed to work directly with the default hand used by the Oculus Quest hand tracking. Currently, this is the most widely used variant of hands for VR and also the one most suited for remote studies due to the large user base of consumers. However, note that the textures can, in principle, be used with other hand models. Assuming the topology is similar (as should be the case for models of human hands) this would only require some changes to the UV coordinates of those models. Additionally, we provide hand textures compatible with the 3D hand models developed for the Oculus controllers; that way researchers can utilize our hand resource for studies using both free hand interaction and controller-based interaction.

We conducted an online and unsupervised study to evaluate the suitability of the hands for remote evaluations, and to get quantitative insights of their effects on embodiment. We evaluate the effects of skin-tone matching as well as owning the non-human hand variants. The study, its hypotheses, method, procedure, and analyses approach was pre-registered ⁴ .

We used a within-subjects design with hand texture as the only independent variable. Hand texture had three levels: skin-tone matched, skin-tone mismatched, and non-human. Participants completed three repetitions of a hand movement task with each hand texture, balanced using a latin square, for a total of nine trials per participant.

From the participants responses we seek to validate the suitability of Hafnia hands for VR research. In particular, when used for embodiment and remote studies, Hafnia hands need to 1) resemble participants’ own hands and hence also elicit a higher sense of body ownership. See Figure 2 for an overview of how participants’ rated their experience with the hands.

As an initial test, we checked whether there were differences between the responses of male and female participants. Fisher’s exact tests show no significant differences between the sexes (Body Ownership: p = 0.5, Agency: p = 0.5, Resemblance: p = 0.7, and Humanness: p = 0.1).

We test the effects of hand representation on body ownership, agency, resemblance, and humanness using Friedman tests for main effects. We use Wilcoxon rank sum tests with Bonferroni correction for post-hoc comparisons.

We find an effect of hand representation on body ownership; χ ² (2) = 109.77, p < 0.001. All pairwise comparisons were significant (p < 0.001). For agency, there was a significant main effect χ ² (2) = 26.27, p < 0.001. Yet, only the alien hand and matched hand were significantly different (p = 0.04). We also found an effect of hand representation on resemblance; χ ² (2) = 150.31, p < 0.001. All pairwise comparisons were also significant (p < 0.001). Finally, we also found a significant effect of hand representation on humanness; χ ² (2) = 141.7, p < 0.001. Post-hoc tests found significant differences between the alien hand and the two realistic hands (p < 0.001).

In summary, the hands provided as part of Hafnia hands showed increased body ownership using a skin-tone matched texture; increased resemblance with a skin-tone matched texture; and decreased humanness with a alien texture. All registered hypotheses were hence found to be true.

These results also validate the suitability of Hafnia hands as a resource in embodiment and remote VR studies. Particularly with respect to body ownership, skin-tone matched hands provide a benefit. Furthermore, we validated that the provided non-human alien hand elicits the opposite response: low body ownership, visual resemblance, and humanness. Hence, the set of provided hand textures covers a wide spectrum of subjective responses, which is required in a range of study designs.

The sample for the evaluation was predominantly young and male, which has previously been shown to cause bias in VR research (Peck et al., 2020). The hand textures visually resemble relatively young human hands, making the sample’s age fit the study; however, future studies could entail textures matched and evaluated for seniors. The hand textures were, however, designed as gender neutral, whereas the evaluation consisted of predominantly males. While we saw no significant differences between male and female responses, this still is a limitation of the evaluation, caused the bias of HMD owners’ demographics. Our participant sample also is biased in terms of skin tone with 96 of them of type II, III, or IV. We hence have only little data from people with very light or dark skin.

As a manually-created resource, Hafnia hands are limited in the range of visual characteristics they cover. In addition to lack of gender and age specificity, the hands also do not account for several other properties, such as: dirtiness, fingernail length/style/color, skin conditions, or finer nuances in skin-tone. A potential solution for this are parametric hand texture models (Qian et al., 2020). However, to make use of the wide range of such models, a parameter acquisition step for each participant is required. The quality of current model outputs also is not consistently at the same level as artist-created textures and does not include non-human textures at all.

While hands are perhaps the most important limb in embodiment research (e.g., the rubber hand illusion), future research resources and studies could consider creating resources enabling full body VR embodiment studies.