This study concerned the perception of bodily orientation. We varied the or
ientation of the body relative to the visible surroundings and, independent
ly, relative to the direction of balance. This created 3 tilt conditions th
at are commonly believed to create conflict between the visual, vestibular,
and somatosensory systems. First, we replicated earlier effects, confirmin
g the existence of the visual frame effect and the Muller effect, when part
icipants made passive judgments about orientation. In later experiments usi
ng the same tilt conditions, participants executed a pointing task. In the
pointing task errors were greatly reduced and in some cases were entirely a
bsent. Based on these data (and similar findings in other studies), we argu
e that illusions of body orientation are highly task-specific and may not b
e general properties of the perception of orientation. The use of tilt rela
tive to different referents made it possible for us to contrast the sensory
conflict interpretation of orientation perception (which predicts frequent
errors arising from indirect perception) with an alternative based on the
pickup of intermodal relations extending across perceptual systems (which p
redicts generally accurate performance arising from direct perception). By
conducting comparisons across pairs of tilt conditions, we were able to hol
d constant the stimulation to individual perceptual systems while varying t
he higher order relation across systems. The results of these pairwise comp
arisons were compatible with the hypothesis that in perceiving orientation
people rely on information in these higher order patterns.