
I wonder what would happen if you were to turn this video into an actual game. Once you become bored or distracted, your attention is likely to fall on the changing maze and you immediately notice what’s going on. Whether you were fooled by the Pac-man video for a long time, or whether you quickly noticed that the background was changing, depends primarily on how focused you were on the Pac-man.

With a well-crafted animation, you can get away with enormous changes! In this case, I probably got away, for a while, with changing pretty much everything. In the video, this is accomplished by having a moving Pac-man, which almost certainly diverts the viewer’s attention away from the background, at least initially.

The reason is that, because of the blank, the whole image changes (from blank to image and vice versa) and the differences no longer correspond to unique changes. However, by adding a blank in between the alternating images, the differences again become difficult to detect. These unique changes grab your attention, and therefore really “pop out”. Now differences correspond to unique changes in the image. This can be very difficult!īut spotting differences becomes very easy when the same two images are presented one after another at the same location. A typical change blindness experiment is similar to a “spot the differences” game, in which two slightly different images are presented side by side, and you have to, well, spot the differences.

This video is a somewhat unusual demonstration of a classic phenomenon called change blindness, which was first described by Ronald Rensink, Kevin O’Regan, and James Clark. How can it be so easy to miss these large changes that are happening right in front of your eyes? And not just a wall being added or removed here and there! The whole maze, with the exception of the tiles around Pac-man, is completely randomized with every new frame. It’s the maze that is continuously changing.

In fact, as you probably noticed after a while, Pac-man is not doing anything weird at all.
