Object recognition can drive motion perception

When two spatially separated spots of light are flashed in rapid succession, apparent motion is seen between them1,2,3. We extended this phenomenon by photographing a face and producing from it a fragmented ‘puzzle picture’ or ‘Mooney face’ in which the face is not initially visible (Fig. 1, left; frame 1) but is seen after 15 to 60 seconds4. Another photograph of the same face seen in profile was used to produce a second Mooney face (Fig. 1, right; frame 2). When the two images were alternated, naive subjects at first saw random, incoherent, two-dimensional (2D) motion between the fragments. But once the face was recognized, it was perceived to rotate unambiguously in three dimensions. We conclude that complex image tokens set up by perceptual learning can drive perception of apparent motion. Figure 1 Two views of a ‘Mooney’-type face produced by digitizing a video image. Left, frame 1; right, frame 2. A demonstration can be produced by showing the two frames alternately as an upside-down movie; nearly all subjects see 2D movement. But if the same movie is viewed upright or the subjects are told it is a face, they invariably see rigid 3D rotation.