Non-informative components of retinal and extra-retinal signals affect perceived surface orientation from optic flow

We test two predictions of a new computational model for the interpretation of the Optic Flow (OF) (Domini, et al., VSS 2012): (1) perceived orientation of a planar surface rotating about an axis coplanar to the surface undergoes a 90° flip whenever a translational component orthogonal to the axis of rotation is added to the OF; (2) the perceptual interpretation of the OF relies on the angular, but not on the linear component of observer’s egomotion. In Experiment 1, a static observer viewed the OF produced by a random-dot planar surface rotating about a vertical axis (e.g., a rigid flag hinging on a vertical pole). This OF induces a veridical perception of the surface orientation. However, consistently with prediction (1), when a vertical translational component was added to this OF, the 3D interpretation underwent a 90° flip (i.e., a rigid flag hinging on a horizontal pole). In Experiment 2, the OFs were actively produced by observer’s head movements. The observer looked at rotating planar surfaces while performing either a lateral translation of the head or a horizontal head rotation (yaw). In one condition the motion of the planar surface was tethered to the motion of the observer, so that the translational component of the OF was nil. In another condition a surface rotating about a static axis of rotation produced the same lateral translation of the OF tested in Experiment 1. Consistently with prediction (2), perceived surface orientation depended on the lateral motion of the OF in the head-translation, but not in the in the head-rotation condition. Experimental results support the model proposed by Domini, et al., (VSS 2012).