In the real world, people are quite accurate at performing visually-directed actions to indicate the distance of floor level targets. However, several studies have shown that there is a compression of these judgments when observers view a simulation of the world using a head-mounted display (HMD). The present study investigated the hypothesis that one cause for this effect is the lack of realism present in the cartoon-like computer graphics that have been used in many of the experiments evaluating space perception in visually immersive displays. Walking to or towards previously viewed targets has been used extensively to evaluate judgments of absolute egocentric distance. The current experiment involved looking at a target, walking while blindfolded in an oblique direction from the original line of sight to the target, and then taking several steps towards the (now unseen) target. The perceived distance is determined based on the original starting point and the intersection of the original line of sight with the final indicated direction. Subjects performed this triangulated walking task after viewing targets at distances of 5m, 10m, and 15m in one of four conditions: (1) the real world, (2) a low-quality virtual world made of a few large planar surfaces and unrealistic texture maps, (3) a wire-frame rendering of the same low-quality virtual world, and (4) a high-quality virtual world constructed from photographic panoramic images. Stereo viewing was used in all four conditions. As expected, distance judgments in the real world were near veridical whereas distance judgments based on low-quality texture mapped computer graphics were significantly compressed. The surprising result was that the amount of compression was nearly the same for the other two computer graphics conditions as well. That is, distance judgments were almost unaffected by the quality of the imagery presented to subjects.