The extrinsic/intrinsic classification of two-dimensional motion signals with barber-pole stimuli.

The perceived direction of different barber-pole stimuli was assessed by adjusting an arrow on the screen. The terminator ratio (TR: number of terminators moving along the long side divided by the number of terminators moving along the small side) was either one or three. In this latter case, the aperture orientation was either vertical or horizontal. The grating was either in the same plane as the aperture (intrinsic condition) or behind the aperture--the frame containing the aperture had a crossed disparity relative to the grating--(extrinsic condition). A nested design with 120 observers was used for the whole study. Five grating orientations were intermingled within any session. With a terminator ratio of three, the results depend strongly on the aperture's orientation. When the rectangular aperture is horizontal, the perceived direction of an intrinsic grating is horizontal (the typical barber-pole illusion), whereas it is only slightly biased towards orthogonal one-dimensional (1D) motion signals (Vp) in the extrinsic condition. When the aperture is vertical, the perceived direction in the intrinsic condition is largely biased toward Vp, and on average it is close to Vp in the extrinsic condition. In this latter case, however, analysing the distributions of responses shows that many responses do not lie around Vp but are clustered near vertical or horizontal. This motion capture depends on the grating's orientation. With a terminator ratio of one, motion capture is present in both the extrinsic and intrinsic conditions. Moreover, a global bias toward horizontal is observed: this horizontal bias is much larger in the extrinsic condition. Altogether, these results suggest that binocular disparity alone is a weak determinant of the extrinsic/intrinsic classification of two-dimensional (2D) motion signals compared to the occlusion cues provided by unpaired regions in binocular images. Second, truly extrinsic 2D motion signals are not suppressed but rather actively compete against each other to capture the 1D motion signals. This results in a perceptual multistability which is much stronger with extrinsic signals. Finally, given the inherent multistability of barber-pole stimuli, high-level factors can alter the strength of this competition and prime any of the 2D motion signals.