Abstract
We introduce a stimulus whose perceptual properties call into question current models of first-order motion processing. This stimulus consists of a diagonal sinusoidal carrier grating with bars drifting up to the left, windowed by a raised, vertical, drifting sinusoid. In slow motion, this stimulus represents a row of barber poles with blurred edges moving either to the right or to the left on a gray background, with diagonal stripes inside the poles moving either upward, downward or not at all, depending on the relative speeds of the carrier and windowing gratings. Foveally, this stimulus can be perceived veridically. However, when viewed peripherally at high temporal frequencies (conditions expected to allow the first order system to dominate the motion percept), this stimulus appears to move as a whole in a consistent, often non-veridical, direction. When the diagonal bars are vertically stationary inside the barber poles, then (as one might expect) this stimulus yields pure horizontal motion. However, if the diagonal bars are moving with even moderate speed inside the barber poles, then the motion evoked is vertical (up or down depending on the direction of motion of bars inside the pole) over a wide range of barber pole speeds. Why this is surprising: (1) Physically, this stimulus translates rigidly in a diagonal (feature-tracking) direction; (2) the vector average of the Fourier components of this stimulus is also non-vertical. Various models of first-order motion processing predict that the motion produced by this stimulus should be in one or another of these two directions. We infer instead that the direction of first order motion is strongly influenced by a "stream-detection" process sensitive to the orientation of contrast-defined stream-beds within which pattern motion is constrained to flow. Remarkably, this process is insensitive to the motions of the stream-beds themselves.
Meeting abstract presented at VSS 2012