An equivalent illuminant analysis of lightness constancy with physical objects and in virtual reality

Several previous studies have found significant differences between visual perception in real and virtual environments. Given the increasing use of virtual reality (VR) in performance-critical applications such as medical training and vision research, it is important to understand these differences. Here, we compared lightness constancy in physical and VR environments using a task where viewers matched the reflectance of a fronto-parallel match patch to the reflectance of a reference patch at a range of 3D orientations relative to a light source. We used a custom-built physical apparatus and four VR conditions: (1) All-Cue (replicating the physical apparatus), (2) Reduced-Depth (no disparity or parallax), (3) Shadowless (no cast shadows), and (4) Reduced-Context (no surrounding objects). Lightness constancy was markedly better in the physical condition than in all four VR conditions. Surprisingly, viewers achieved a degree of lightness constancy even in the Reduced-Context condition, despite the absence of lighting cues. In a follow-up experiment, we re-tested the All-Cue and Reduced-Context conditions in VR with new observers, each participating in only one condition. Here, we found lower levels of constancy than in the first experiment, suggesting that experience across multiple experimental settings and possibly exposure to the physical apparatus during instructions had enhanced performance. We conclude that even when robust lighting and shape cues are available, lightness constancy is substantially better in real environments than in virtual environments. We consider possible explanations for this finding, such as the imperfect models of materials and lighting that are used for rendering in real-time VR.