What's moving? You? The background? Something inbetween?

When I was 12 years old, I sometimes got to ride the train from Seattle to my aunt's house in Portland. Staring at the countryside flashing past the train window, it seemed to me that the landscape was rotating in a giant circle: Nearby objects flashed past the train as expected -- they appeared to move the opposite way the train was going. But the mountains in the distance seemed to be moving forward, faster than the train. It was as if the land next to us was just a vast turntable, rotating rapidly as we stayed in the same place. This video (not my own) captures some of the effect:

I knew this must be an illusion caused by the motion of the train, but I could never satisfactorily explain to myself why or how it happened. What I didn't realize is that scientists had been measuring this illusion, called the Filehne Effect, for over fifty years. A complementary effect, the Aubert-Fleischl effect, has been known for more than a century. In the Aubert-Fleischl illusion, when we track an object moving across a stationary background (like a train passing in the distance), we underestimate its speed.

It's quite difficult to measure these effects, because they involve relative motion. How fast is the train or the background apparently moving? You don't know; you just know that they seem to move.

One novel method was recently demonstrated by Richard Dyde and Laurence Harris. It involved constructing an elaborate apparatus:

Volunteer observers lay on a movable platform and look up at the ceiling. A partially reflective mirror is mounted at an angle above their head. The mirror allows them to see a small disk apparently hovering above them, below the ceiling, but also to see through to the ceiling above. The disk is actually produced by a computer display to the viewers' side, outside of their view. This side view shows their motion on the platform:

As they move back and forth, the computer tracks the motion of the platform and adjusts the disk's display to match the viewer's motion. The question is, how much do you have to move the disk in order to make it seem stationary? It's not a difficult calculation, once you know what perspective to take. There are three possibilities:

1. The disk is not moving at all -- it stays in exactly the same place and the viewer moves
2. The disk appears to stay in the same spot relative to the background. To do this, it must move, but not as much as the viewer moves
3. The disk moves exactly as much as the viewer, "hovering" exactly over their head

Dyde and Harris programmed the disk to move systematically at increments between (1) and (3) above. Viewers watched while either moving the platform themselves or being moved by an experimenter, in time with a metronome. With their free hand they clicked on a mouse to indicate if the disk was moving along with them, or moving the opposite way from themselves. In addition, the experiment was conducted both with the lights on (so the ceiling was visible) or off.

Only (1) is the "right" answer -- the disk is actually stationary. But it might seem stationary with respect to the viewer in either case (2) or (3). In fact, viewers weren't given "stationary" as an option -- they could only judge when the object appeared to be moving with them or against them. Here are the results for when the viewers moved the platform themselves:

The horizontal axis shows the "gain" of the object relative to the viewer. A gain of 1.0 corresponds to the geometrically correct movement -- when the disk is literally not moving. A gain of 0 is when the disk is moving exactly along with the viewer.

Since viewers weren't given the choice of saying the disk was stationary, the point at which half of the disks were said to be moving with the viewer and half were moving opposite the viewer can be estimated as when it seemed about stationary.

Whether or not the room was dark, viewers made systematic errors -- they responded as if it was not moving when it was actually moving quite a bit. In nearly every case when they said it was moving against them, it was actually still moving with them. They made significantly larger errors when they were moving in the dark, and there was no significant difference when they were moving themselves versus being moved by the experimenter.

While being able to see the background reduces the size of the error, the error occurs in both situations, suggesting that we are, in fact, tremendously bad at judging relative motion.

Back to my train-riding example from childhood, there was a point in the center of my visual field that appeared to not be moving at all compared to me. Of course it was: I was careening along at 70 miles per hour. But it seemed to be standing still -- or rather, I seemed not to be moving relative to it. Fortunately I was, or I never would have gotten to my Aunt's house!

Richard T. Dyde, Laurence R. Harris (2008). The influence of retinal and extra-retinal motion cues on perceived object motion during self-motion Journal of Vision, 8 (14), 1-10 DOI: 10.1167/8.14.5