Visual, vestibular and body cues to upright are weighted in proportion to their reliability
Poster Presentation

Laurence Harris
Psychology, York Univ.

Richard Dyde
Centre for Vision Research, York Univ.

Michael Jenkin
Dept. Computer Science and Engineering, York Univ.

     Abstract ID Number: 83
     Full text: PDF
     Last modified: June 30, 2005

Abstract
We measured the perceptual upright whilst manipulating the orientation of the visual background and posture of the observer to assess the relative contributions of the body axis, visual background and gravity. Subjects made forced-choice decisions identifying the character ‘p’ shown in various orientations as either a 'p' or a 'd'. The orientations of maximum ambiguity were derived from psychometric functions of the probability of choosing one or other identity as a function of the character’s orientation. The perceptual upright was defined as midway between these two orientations. The perceptual upright was modeled as the vector sum of the orientations of the body, visual background and gravity with weightings of 2.6: 1.2: 1 respectively. The variances of each cue were obtained from the psychometric functions for conditions where the body only (supine with no background), body and gravity (upright with no background) and all three cues (upright with upright background) contributed. The ratio of the reliability of each cue derived from these variances agreed well with the ratio of the weightings obtained from the vector model. This correlation is in line with models which propose that the reliability of a cue is predictive of its importance in multisensory integration.