| |
Abstract:
Abstract: Our ability to recognize objects across changes in
viewpoint has been commonly studied with stationary subjects and
static displays of isolated single objects. In contrast, real-world
recognition involves ambulant observers in cluttered environments
containing multiple objects. Moreover, viewpoint changes may be
generated through a rotation of the object (ROTATION) or through
movement of the observer (WALKING) - conditions that preserve
different relations between the viewer-, object-, and
environment-centered frames of reference in a scene and that
provide different information about how the environment is
changing. Using virtual reality (VR) to create a dynamically
changing world responsive to a subject's movements, we compared
recognition of novel 3D objects across viewpoint changes generated
by ROTATION, WALKING, or a condition in which subjects were
instantaneously transported to a new position (TRANSPORT).
Critically, in all conditions the same visual information was
available from each object view, therefore differences in
performance can be attributed to differences in the relations
between reference frames and extra-retinal information such as
odometry. There were, however, no strong differences in the pattern
of performance across the different conditions - in each case
larger changes in viewpoint relative to the object yielded
proportionally slower responses and greater errors. Such results
indicate that even in conditions where observers consciously
generate a change in viewpoint, they are unable to utilize
view-invariant recognition mechanisms.
|
