Virtual reality (VR) is increasingly being used as a research platform for investigating human responses to environmental variables. While VR provides tremendous advantages in terms of variable isolation and manipulation, and ease of data-collection, some researchers have expressed concerns about the ecological validity of VR-based findings. In the current study we replicated a real-world, multi-level educational facility in VR, and compared data collected in the VR and real-world environments as participants (n=36) completed identical wayfinding tasks. We found significant differences in all of the measures used, including distance covered, number of mistakes made, time for task completion, spatial memory, extent of backtracking, observation of directional signs, perceived uncertainty levels, perceived cognitive workload, and perceived task difficulty. We also analyzed potential age-related effects to look for heightened VR/real response discrepancies among older adult participants (>55 years) compared to younger adults. This analysis yielded no significant effects of age. Finally, we examined the spatial distribution of self-reported wayfinding uncertainty across the building floorplan, finding that areas in which uncertainty was most pronounced were similar between the real-world and VR settings. Thus, participants appeared to be responding to the same environmental features in the real and VR conditions, but the extent of these responses was significantly different. Overall, the findings suggest that when VR is used to contrast varying environmental design conditions the resulting data should be interpreted cautiously and should not be generalized into real-world conclusions without further validation.