In previous work, we have found significant differences in participants' distance perception accuracy in different types of immersive virtual environments (IVEs). Could these differences be an indication of, or consequence of, differences in participants' sense of presence under these different virtual environment conditions? In this paper, we report the results of an experiment that seeks further insight into this question. In our experiment, users were fully tracked and immersed in one of three different IVEs: a photorealistically rendered replica of our lab, a non-photorealistically rendered replica of our lab, or a photorealistically rendered room that had similar dimensions as our lab, but was texture mapped with photographs from a different real place. Participants in each group were asked to perform a series of tasks, first in a normal (control) version of the IVE and then in a stress-enhanced version in which the floor surrounding the marked path was cut away to reveal a two-story drop. We assessed participants' depth of presence in each of these IVEs using a questionnaire, recordings of heart rate and galvanic skin response, and gait metrics derived from tracking data, and then compared the differences between the stressful and non-stressful versions of each environment. Pooling the data over all participants in each group, we found significant physiological indications of stress after the appearance of the pit in all three environments, but did not find significant differences in the magnitude of the physiological stress response between the different environment conditions. However, we did find significant differences in the change in gait: participants in the photorealistic replica room group walked significantly slower, and with shorter strides, after exposure to the stressful version of the environment, than did participants in either the photorealistically rendered unfamiliar room or the NPR replica room conditions.