Perceiving motion

Another key perceptual quality to consider in exploring how the visual brain works is motion. Although Zhiyong Yang had struggled early on to make some sense of motion in empirical terms, we had put this issue aside while focusing on brightness, color, and geometry as perceptual domains that would be more tractable in trying to understand how brains use sense information that can't specify the physical world directly. The main reason for this reticence was that we had no idea how to determine object motion in three-dimensional space, the information we needed to test the idea that perceptions of motion arise empirically. A less straightforward concern was the conceptual difficulty that an empirical explanation of motion entails. It was hard to imagine that successful behavior in response to moving objects could possibly be based on the frequency of occurrence of retinal stimulus sequences. Most vision scientists take it for granted that the brain generates perceptions of motion and visually guided actions using the features in retinal images to compute object motion "online." Even for us, the idea that the perception of motion and the complex actions involved in a motor response such as catching a ball could be based on trial-and-error behavior seemed bizarre.

In physical terms, motion refers to the speed and direction of objects in a three-dimensional frame of reference and follows Newton's laws. In psychophysical terms, however, seeing object speeds and directions is defined subjectively. For example, we don't see the motion of a clock's hour hand or a bullet in flight, even though both objects move at physical rates that are easily measured. The range of object speeds projected onto the retina that humans have evolved to see as motion is from roughly 0.1° to 175° of visual angle per second, a degree of visual angle being 1/360 of an imaginary circle around the head (about the width of a thumbnail at arm's length). Below this range, objects appear to be standing still. As speeds approach the upper end of the range, moving objects begin to blur and then, as with the bullet, become invisible.

One of the things that had puzzled natural philosophers and neuroscientists thinking about motion is the obvious way that context affects motion perception. Depending on the circumstances, the same speed and direction projected onto the retina can elicit very different perceptions. Although such phenomena were often treated as special cases or were simply ignored, if what we had been saying about vision was true, then the perception of motion (including these anomalies) should have the same empirical basis as lightness, brightness, color, and perceptions of geometry. All the motions we see should be understandable in terms of the biological need to contend with the inverse problem as it applies to moving objects.

How the inverse problem pertains to seeing motion is easy enough to understand. For obvious reasons, observers must respond correctly to the real-world speeds and directions of objects, and these responses are certainly initiated by the speeds and directions of objects that determine stimulus sequences projected onto the retina. But as illustrated in Figure 12.1, when objects in three-dimensional space are projected onto a two-dimensional (2-D) surface, speed and direction are conflated in the resulting images. As a result, the sequence of actual positions in 3-D space that define motion in physical terms is always uncertainly represented in the sequence of retinal positions that moving objects generate.

If contending with this problem depended on the empirical framework that we had used to rationalize other visual qualities, the perceptions of motion elicited by image sequences should accord with—and be predicted by—the frequency of occurrence of the retinal image sequences generated by moving objects in the world. However, testing this idea was not so easy. The most formidable obstacle was acquiring a database of the 2-D projections arising from the speeds and directions of objects moving in 3-D space. Although data relating retinal projections to real-world geometry had been relatively easy to obtain for static scenes using laser range scanning (see Chapter 11), we had no technical way to get the information about the direction and speed of objects moving in space that we needed to determine the frequency of occurrence of speeds and directions on the retina.

Figure 12.1 The inverse optics problem as it pertains to the speed and direction of moving objects projected onto the retina. Objects (black dots) at different distances moving in different trajectories (arrows) with different speeds can all generate the same 2-D image sequence on the retina (the diagram shows a horizontal section through the right eye). Therefore, speeds and directions in retinal images cannot specify the speeds and directions of the real-world objects that observers must deal with. (From Wojtach, et al., 2008. Copyright 2008 National Academy of Sciences, U.S.A.)

Nevertheless, we could approximate human experience with object motion using a virtual world in which moving objects were projected onto an image plane (a stand-in for the retina) in a computer simulation (Figure 12.2). Although grossly simplified with respect to all the empirical factors that influence natural motion, this surrogate for experience with moving objects was not unreasonable. Most important, it accurately represented perspective, the major determinant of the difference between object motion in 3-D space and the speeds and directions projected onto the 2-D retinal surface (see Figure 12.1). By sampling the image plane in the virtual environment, we could tally up the frequency of occurrence of the projected speeds and directions arising from 3-D sources whose speeds and directions were known. We could then use this approximation of motion experience to predict the perceived speeds and directions that should be seen in response to various motion stimuli if motion perception is generated empirically.

As in every other perceptual domain, there are puzzles in motion perception whose bases have been debated for decades. Two specific examples that people have struggled to explain are the flash-lag effect, which concerns the perception of speed, and the effects of occluding apertures, which concern the perception of direction (recall that speed and direction are the two basic characteristics of perceived motion). Two postdocs, Kyongje Sung and Bill Wojtach, took on the challenge of explaining these effects in empirical terms. Kyongje had earned his Ph.D. at Purdue in 2005 studying how people carry out visual search tasks (looking for a specific object in a scene), and he was the first card-carrying psychophysicist to join the lab. Bill had gotten his Ph.D. at about the same time in philosophy at Duke working on perception; he had come to lab meetings while working on his doctorate and eventually decided to pursue a career that tapped into both philosophy and neuroscience. Wojtach and Sung made a somewhat unlikely scientific pair, but they complemented each other's skills nicely and eventually succeeded in making the case that perceptions of motion are indeed based on the same strategy as other visual qualities, and for the same general reasons.

They first explored the flash-lag effect, a phenomenon that had been studied since the 1920s without any agreement about its cause. When a continuously moving stimulus is presented in physical alignment with an instantaneous flash that marks a point in time and space, the flash is seen as lagging behind the moving stimulus (Figure 12.3A). Moreover, the faster the speed of the stimulus, the greater the lag ( Figure 12.3B). The effect is actually one of several related phenomena apparent when people observe stimuli in which a moving object is presented together with an instantaneous marker. For example, if the flash occurs at the start of the trajectory of an object, the flash is seen as displaced in the direction of motion (the so-called Fröhlich effect). And if observers are asked to specify the position of a flash in the presence of nearby motion, the flash is displaced in that direction (the flash-drag effect). People don't ordinarily notice these discrepancies, but they are quite real and reveal a systematic difference between the speed projected on the retina and the speed we see. This difference raises questions about how the perception of speed is related to behavior and why the discrepancies exist in the first place. Despite various proposed explanations, there were no generally accepted answers.

Figure 12.2 Determining the frequency of occurrence of image sequences generated by moving objects in a virtual environment. Diagram of a simulated visual space (red outline) embedded in a larger spherical environment; objects moving randomly in different directions at different speeds entered the visual space and projected onto an image plane (blue outline). By monitoring the projections, we could determine the frequency of occurrence of the speeds and directions in image sequences generated by objects moving in 3-D space. (After Wojtach, et al., 2008. Copyright 2008 National Academy of Sciences, U.S.A.)

Figure 12.2 Determining the frequency of occurrence of image sequences generated by moving objects in a virtual environment. Diagram of a simulated visual space (red outline) embedded in a larger spherical environment; objects moving randomly in different directions at different speeds entered the visual space and projected onto an image plane (blue outline). By monitoring the projections, we could determine the frequency of occurrence of the speeds and directions in image sequences generated by objects moving in 3-D space. (After Wojtach, et al., 2008. Copyright 2008 National Academy of Sciences, U.S.A.)

Figure 12.3 The flash-lag effect. A) When a flash of light (indicated by the asterisk) is presented on a screen in physical alignment with moving stimulus (the red bar), the flash is perceived to be lagging behind the position of the moving object. B) The apparent lag increases as the speed of the moving object increases. C) The amount of lag seen as a function of object speed, determined by asking subjects to adjust the position of the flash until it appeared to be in alignment with the moving stimulus. (After Wojtach, et al., 2008. Copyright 2008 National Academy of Sciences, U.S.A.)

percept (slow speed)

*

Was this article helpful?

0 0

Post a comment