Failures of depth magnitude estimation in virtual but not physical stimuli
Poster Presentation: Tuesday, May 20, 2025, 8:30 am – 12:30 pm, Pavilion
Session: 3D Processing: Space, coordinate frames, virtual environments
Schedule of Events | Search Abstracts | Symposia | Talk Sessions | Poster Sessions
Arleen Aksay1 (), Laurie M. Wilcox1; 1York University, Toronto, Canada
When asked to estimate depth magnitude in naturalistic, cluttered stimuli observers lacking experience with virtual reality (VR) exhibit little to no scaling of depth with binocular disparity. Here, we asked whether this failure reflects a general inability to generate metric depth estimates for these types of stimuli or if it is related to cue conflicts that limit the use of vergence to estimate viewing distance in VR. To this end, we rendered low (‘branch’) and high (‘thicket’) complexity stimuli in VR and 3D printed exact physical replicas. In the branch condition, two branches were presented on either side of a central reference branch. The thicket condition consisted of two mirrored clusters of overlapping branches centred around the reference. The total disparity ranged from 0.33 to 2.10 degrees. Separate groups of novice observers estimated the 3D volume of branches (N=28) or thickets (N=25) using a virtual ruler. Both types of stimuli were tested in VR and physically; the physical stimuli were tested monocularly and binocularly. Our VR results replicated previous outcomes: novice observers showed limited depth scaling irrespective of complexity. Their binocular functions were relatively flat and similar to those obtained monocularly. Very different results were obtained using physical stimuli depth, for which estimates scaled well with disparity. These results support our hypothesis that novice observers’ poor depth scaling is due to unreliable distance information in VR, which is needed to scale binocular disparity. Importantly, experienced observers show reasonable depth scaling for virtual stimuli, suggesting that they are able to ignore cue conflicts. These findings have important implications, both for cue integration studies, which typically involve experienced observers, and for the use of VR in studies of depth perception.
Acknowledgements: Natural Sciences Research Council of Canada (NSERC); CF-REF program Vision Sciences to Applications (VISTA)