Motion-induced distortion of shape

Abstract

Motion, position, and form are intricately intertwined in perception. Motion distorts visual space, resulting in illusory position shifts such as flash-drag and flash-grab effects. The flash-grab displaces a test by up to several times its size. This lets us use it to investigate where the motion-induced shift operates in the processing stream from photoreceptor activation to feature activation to object recognition. We present several canonical, highly familiar forms and ask whether the motion-induced shift operates uniformly across the form. If it did, we could conclude that the effect occurred after the elements of the form are bound. However, we find that motion-induced distortion affects not only the position, but also the appearance of briefly presented, canonical shapes (square, circle, and letter T). Features of the flashed target that were closest to its center were shifted in the direction of motion more than those further from its center. Outline shapes were affected more than filled shapes, and the strength of the distortion increased with the contrast of the moving background. This not only supports a nonuniform spatial profile for the motion-induced shift but also indicates that the shift operates before the shape is established, even for highly familiar shapes like squares, circles, and letters.

Figure 1.

Spatially nonuniform shift. If the flash grab effect is larger at the contrast border than immediately to its left and right, the central stem of the T will be shifted leftward relative to the endpoints of the top bar, producing an asymmetrical shape.

Figure 2.

Target shapes and their adjustable versions. Participants adjusted the test form between the two extremes until the flashed test appeared to match the reference shape. The middle column shows the reference shapes, and these were scored 0 (no illusion) if participants reported that this shape matched the reference shape. The left column shows the exaggerations of the illusory distortions seen in the pilot study and were scored −1 if chosen. The right column shows the reverse of the left column, and these were scored +1, representing a very strong illusion that required an extreme opposite distortion to cancel.

Figure 3.

Trial sequence. Each trial started with the presentation of the target shape. The stimulus was then presented in motion, with the adjustable target briefly flashed on every second reversal.

Figure 4.

Distortion of rectangular shapes as a function of contrast. The top row represents the shape that was perceived as a square in each condition. Error bars represent ±1 SEM.

Figure 5.

Distortion of elliptical shapes as a function of contrast. The top row represents the shape that was perceived as a circle in each condition. Error bars represent ±1 SEM.

Figure 6.

Distortion of the T as a function of contrast. The top row represents the shape that was perceived as a symmetrical T in each condition. Error bars represent ±1 SEM.

Figure 7.

The central part of the shift data from Cavanagh and Anstis (2013) is shown on the left, converted to dva. Their test probe was a compact green flash, but our T-shape is shown here at the appropriate size. Notice that the shift values would be larger at the end points of the T than at its stem. On the right, the effect of these different shifts is shown for the T-shape on top and the circle on the bottom. The distortions in the experiment are in the directions predicted here but larger.