The induced motion effect is a high-level visual phenomenon: Psychophysical evidence

Abstract

Induced motion is the illusory motion of a target away from the direction of motion of the unattended background. If it is a result of assigning background motion to self-motion and judging target motion relative to the scene as suggested by the flow parsing hypothesis then the effect must be mediated in higher levels of the visual motion pathway where self-motion is assessed. We provide evidence for a high-level mechanism in two broad ways. Firstly, we show that the effect is insensitive to a set of low-level spatial aspects of the scene, namely, the spatial arrangement, the spatial frequency content and the orientation content of the background relative to the target. Secondly, we show that the effect is the same whether the target and background are composed of the same kind of local elements-one-dimensional (1D) or two-dimensional (2D)-or one is composed of one, and the other composed of the other. The latter finding is significant because 1D and 2D local elements are integrated by two different mechanisms so the induced motion effect is likely to be mediated in a visual motion processing area that follows the two separate integration mechanisms. Area medial superior temporal in monkeys and the equivalent in humans is suggested as a viable site. We present a simple flow-parsing-inspired model and demonstrate a good fit to our data and to data from a previous induced motion study.

Keywords: higher-order motion; local motion; models; motion; neural mechanisms; optic flow; perception; perceptual organization; scene perception.

Figure 1.

Illustration of the flow parsing hypothesis (Warren & Rushton, 2007). Based on an image in Warren and Rushton (2009). Obtaining the world-centered velocities of the background and object (right panel) from the retinal velocities (left panel) involves subtracting the velocities created by self-motion through the environment, that is, adding the negative of the velocity field created by self-motion (central panel). Note that the oblique retinal motion of the ball is perceived (correctly) as vertical after parsing out self-motion.

Figure 2.

Example stimuli used in our experiments. (A) A representation of the “1D” stimulus used in our first experiment; target ring against a field background. Shown is the central portion of the stimulus display plus dashed lines to indicate the shape of and direction of motion for the target (red) and the background (blue). All Gabor envelopes were stationary but each sinusoidal carrier drifted at a speed consistent with the overall motion of the target or background to which it belongs. Although the actual target object velocity is upwards it appears to move up and to the left. Please see video 2A in online Supplemental Material. (B) An example “2D” stimulus used in the fourth experiment; ring target with ring “background” where the background is inside the target. Each small patch stayed in place but the plaid patterns within them drifted so as to evoke a separate sense of motion in the target and background. In the video associated with the image, the outer target ring velocity is upwards, but it appears to move up and to the left as the inner background ring has a rightward component to its velocity. Please see video 2B in online Supplemental Material.

Figure 3.

Experiment 1 part 1; background field versus background ring. Shown is target direction that is perceived as vertical for a range of background directions for both the background ring and background field conditions. Both are measured in degrees clockwise of vertical. This convention is used in all graphs to follow. The averaged performance of the group of observers is plotted. Error bars represent 1 SD.

Figure 4.

Experiment 1 part 2 and experiment 2; background ring inside versus outside of the target ring with high and low spatial frequency (“sf”) local elements. The target direction normalization process is described in the text. Error bars represent 95% confidence intervals.

Figure 5.

Comparison of our results with those of Farrell-Whelan et al. (2012) and Marshak and Sekuler (1979)—the first being an induced motion experiment and the second a direction repulsion experiment. Note that both the Target Direction and Background Direction are measured relative to the perceived target direction. The dashed arrow points to the peak of the Marshak curve and the full arrow points to the peak of the Farrell-Whelan curve. Pilot studies indicated that our repulsion effects peaked in the same place as Farrell-Whelan. Peaking at 90° is a feature of induced motion (see text). Data taken from Experiment 1 above, (Farrell-Whelan et al., 2012) and (Marshak & Sekuler, 1979).

Figure 6.

Experiment 4; comparing background and target types—1D and 2D—along with background relative position. The target direction normalization process is described in the text. Error bars represent 95% confidence intervals.

Figure 7.

Simple flow-parsing inspired model applied to our data and that of Farrell-Whelan et al. (2012) (F-W). See text for details.