Abstract

An observer who approaches a planar surface that rotates about the vertical axis (e.g. a flag hinging on a pole) generates the same Optic Flow (OF) that is produced by a planar surface that rotates about an horizontal axis viewed by a static observer. In spite of this ambiguity, perceived surface orientation by the active observer is usually veridical. This result is consistent with an interpretation of the OF that takes into account egomotion signals (1). Here, we suggest an alternative interpretation based on a computational model that ignores linear egomotion signals (2, 3). An implication of our model is that perceived orientation should flip by 90 degrees whenever the OF undergoes a translational motion in a direction orthogonal to the surface axis of rotation. OF translational motion is always present when an observer moves towards or away from a stimulus display, due to the natural rotations and translations of the head. Main conclusion: In the present experiments, we tested our alternative explanation by asking observers to judge surface orientation in three conditions: 1) when a random-dot planar surface is rotated about a stationary axis, 2) when the axis of rotation was tethered to a coordinate system centered on the observer’s head, so as to eliminate the translational components of the OF, and 3) when a translational component was added to the OF produced in 2). The results are consistent with the predictions of our model. Perceived surface orientation (i) was veridical in 1), (ii) was ambiguous in 2), and (iii) underwent a 90 degrees flip with respect to veridical in 3). A similar pattern of results was found when the same OFs, generated by the observer’s movements, were replayed to a static observer. Next steps: This is inconsistent with the idea that extra-retinal information resulting from head movements is used to extract a veridical interpretation of optic flow.