Our ability to see change over time

Take a look at the following movie (quicktime required). The movie will alternately flash a picture of a desk and a patterned block. Your job is to see if anything about the picture of the desk changes each time it flashes. Don't replay the movie when you get to the end; just stop.

Did you notice any changes? Most people won't spot any changes at all when they watch this movie the first time. But watch the image as you press play again, and you'll see that the desk has changed significantly from the beginning to the end of the movie. I actually rotated it by two degrees at each point along the way. You can drag the handle underneath the movie to fast-forward through the whole thing and convince yourself that the desk was smoothly rotating the entire time.

Andrew Hollingworth of the University of Iowa and John M. Henderson of Michigan State discovered this phenomenon and explored it in "Sustained Change Blindness to Incremental Scene Rotation: A Dissociation Between Explicit Change Detection and Visual Memory" (Perception and Psychophysics, 2004). They found that even though people could easily distinguish between the beginning and end positions of the movie, people can't recognize the change when it is done as a series of small changes (People are able to recognize even small rotations when images are displayed without a "pattern mask" inbetween. By using the pattern mask, we can test people's memory for the position of the objects. Detecting motion is a separate perceptual task from visual memory).

So what is the nature of our visual memory? We clearly remember something about the scene we are shown: we know it's a desk, for example. Perhaps what we're doing is remembering the first position and then comparing it to each subsequent position; the reason that we don't recognize the change each time is because the memory degrades over time. Hollingworth and Henderson devised an experiment to test that question. They showed movies where the desk rotated as in the movie above (but with a nicer picture, and using 1-degree increments). When the desk reached a rotation of 20 degrees from the original position, the pattern mask was flashed one more time, and then the desk was returned to the original position. Participants immediately noticed a change. So clearly, they weren't simply comparing each image to the original picture: from their perspective, what seemed to happen is they were watching the same image over and over again, and then it suddenly shifted 20 degrees counterclockwise.

Somehow, as the image was shifting, so—unconsciously—was the viewers' memory of the image. Why are we unable to detect these shifts in our memory? Imagine a world in which we could. Every time we turned our heads, even slightly, we would seem to see the world changing around us. This, of course, would be a false perception: most of the time, the world is not moving, we are. So we are better off assuming that the world is stable.

There are some interesting connections between this research and the Whitney study I discussed yesterday. Clearly a lot of our visual system is adapted to handle the fact that we are constantly moving through a basically stable world. Both of these studies show that there is a tremendous amount of visual information that our minds simply ignore, in order to build a more coherent picture of the world.