Abstract
We have found that when a virtual room expands around an observer as they move through it, binocular and motion parallax cues fail to override the observer's assumption that the room is static, leading to gross errors in size constancy. Here, we investigated whether this assumption could be changed.
Wearing a head mounted display in an immersive virtual reality system, subjects compared the size of an object visible only on the left side of a room (when the room was small) with one visible on the right (when everything in the room was 4 times larger and 4 times more distant) [condition A]. Subjects then repeated the experiment in a similar expanding room that contained constant-sized objects (human figures) [condition B]. Here, size judgements were much closer to veridical. After training, the improved size judgements persisted even without the human figures [A,B,A], showing subjects' prior assumptions had been changed.
Other subjects repeated the experiment using a training stimulus in which texture elements on the walls and floor remained a constant physical size as the room expanded [condition C]. This gave a much stronger impression that the room changed size and more accurate size matches. However, it was less effective in changing subjects' size matches in the test phase [A,C,A]. Direct comparison in the same subjects [A,C,A,B,A] confirmed the greater effectiveness of condition B. Conclusion: Breaking subjects' assumption of a stable world is most effective when visual stimulation during the training phase is similar to that during the test phase.
Supported by the Wellcome Trust and Royal Society, UK