Cognitive Studies/Psychology/Visual Studies 201:

CAVE Virtual Reality Experiment

Laboratory Module

by Douglas R. Elrod, Cornell University


How do we maintain our upright posture? One might suspect that we simply sense the direction of gravity, and react to that. However, it is generally true that humans, and other animals, have other sources of information about how they are standing, such as the visual flow of the environment around them. And it has been found that if there is relevant information that can be picked up, it is generally not ignored, but used.

In the experiment that is described here, we do a simplified version of Stoffregen (1985), which involved slowly moving the walls of a room containing the experimental participant, and measuring how they move to compensate. The idea is that slow oscillations in the environment can be mistaken for the inherent variability in the maintenance of one's own position, and corrective movements will then be initiated to counter these motions.

We investigate this in the virtual environment of the Cornell CAVE, where we can change how the "environment" moves, in systematic ways, and observe how this affects stance-maintaining behavior.

What is the CAVE?

A CAVE is a "Cave Automatic Virtual Environment". Cornell University's CAVE is described at It projects a visual environment onto two walls and the floor, which together form the corner of a 10'x10'x10' cube. By wearing stereoscopic glasses with position sensors attached, subjects can interact with virtual objects, which appear to be in the middle of the CAVE.

Reading Assignment

Laboratory Exercise

The following laboratory exercise examines how we use information from the environment to correct our stance (the way we are standing). In particular, it compares information from the radial flow (as we approach an object) of the visual field with that provided by the lamellar flow (as we pass by an object). It is adapted from Stoffregen (1985), to use the virtual reality environment of the Cornell CAVE.


When a participant runs in the experiment, he or she will put on the stereoscopic glasses with the position-tracker attached. Then the participant should go to the designated position in the CAVE and stand there, looking straight ahead, in the direction of the corner where two of the video walls meet (which is in the middle of the far wall of the virtual room, see Figure 1).

Moving room, from viewpoint
Figure 1. Moving room, seen from the participant's viewpoint

The position to stand is located so the corner of the CAVE is directly ahead, with the left and right walls equidistant, and the left and right edges of the CAVE just out of view when one looks straight ahead. This maximizes the visual angle of the virtual room.

Please try to keep looking straight ahead, and maintain an upright posture, for the duration of the experiment.

Each trial takes one minute. During that time, the "virtual walls" projected on the screens may be subtly moving, or not, in particular directions.

Each participant will run in nine randomly-ordered trials, three of which have the room moving forward and back, in a sinusoidal motion, three of which have the room moving side to side, again sinusoidally, and three of which have the room stationary. So the experiment should take about 10 minutes per participant.


Data Collected

As mentioned above, there are three types of trials: Those with no motion, those with motion parallel to the line of sight, and those with motion perpendicular to the line of sight. This is the independent variable of the experiment.

The dependent variables, measured by the tracker attached to the stereoscopic glasses, are three reporting head position (x, y, z) and three for head orientation (α, β, γ). The CAVE defines -z as a constant elevation line from the center of the CAVE to the corner of the CAVE, +x as the direction to the right when facing -z, and +y as up. Note that -z is always the direction of gaze in this experiment.

The position of the virtual room at each time is also available, for comparison against the head position data.

Note: It's best to edit a copy of the data file, rather than the original data file itself, given the risk of deleting data!

Data Analysis

Stoffregen (1985) looked at the pattern of correlations between room movement and head movement in the direction of room movement. You can do a similar analysis by doing a linear regression with head position (in the direction of room movement) as the dependent variable. In the forward-back trials, this corresponds to movement along the CAVE's z-axis, and in the side-to-side trials, this corresponds to movement along the CAVE's x-axis. The independent variable would then be the position of the room, as it moves along that same axis.

We expect to find a positive correlation between these two variables, as the participant attempts to keep the same position relative to the room. This would result in a significant positive linear coefficient in the regression.

It would be useful to compare the magnitude of the linear regression coefficient between the forward-back condition and the side-to-side condition. One way to do this is a regression combining the forward-back trials and side-to-side trials, with the dependent variable, HeadPosition, being the head-position coordinate appropriate to the type of trial (z-axis for forward-back trials, and x-axis for side-to-side trials). Then define an independent variable RoomPosition as the value of the appropriate room-position coordinate (z-axis for forward-back trials, and x-axis for side-to-side trials). One would then introduce two "dummy variables", "ForwardA" which is 1 for forward-back trials and 0 for side-to-side trials, and "ForwardB" which is the z-axis room-location value for forward-back trials, and 0 for side-to-side trials. The purpose of ForwardA is just to account for differences in the participants' initial positions along the forward-back and side-to-side dimensions; so we ignore its regression coefficient. If the regression coefficient for ForwardB is significantly different from zero, then people are more sensitive to one form of motion than the other. A significant positive value for this coefficient would indicate greater sensitivity to forward motion (radial flow) and a significant negative coefficient would indicate greater sensitivity to side motion (lamellar flow).

Other possible analyses: Thought Questions: