World Cup Soccer and visual memory

In many ways, my career has been dominated by efforts to make "work" bear as much resemblance to "having fun" as possible. Today's article only confirms that rule. Yesterday afternoon, I spent an hour watching a World Cup soccer match, and for once I could claim that it was completely relevant to work. I could still remember Diego Maradona's stunning 1986 run through the center of the English defense as vividly as if it happened yesterday, and was hoping to form a similarly vivid memory. Unfortunately, yesterday's lackluster 0-0 tie between France and Switzerland didn't provide me with anything remotely approaching that experience.

So what does all this have to do with cognitive psychology? If you study visual memory, quite a bit. One problem that's interested researchers for decades is the question of how visual memories are represented. Do we encode the 3-D event in memory, so that if we saw it from a different angle our recollection would be just as good? Or do we encode it only from the vantage point we saw? Soccer matches offer a perfect means of studying this phenomenon, since most of us watch on TV, with a single camera view.

A group led by Bärbel Garsoffky developed a method of testing these questions. They showed German college students videos of key goal-scoring plays from soccer matches in the 1988 World Cup. They chose only games during which one team scored at least two goals; the students were only shown a video of one of the goals. Then they were tested with still photos. Eight of the still shots were were from the play they had watched, and eight were from the other goal-scoring play during the same match. The students had to indicate whether each still image was part of the play they had watched. Two of the eight "correct" photos were taken from the same vantage point, but the other six were taking from different angles -- either 45, 90, or 135 degrees away from the initial view. Here are the results:

As the still photos were rotated farther away from the initial viewing angle, the students were less and less accurate. The researchers also gave participants a short quiz about their soccer expertise: those who watched fewer than 1,000 minutes of soccer during the previous year were "novices," and students who watched more than 2,000 minutes were "experts" (the average for experts was over 7,000 minutes--more than 70 matches!). Soccer expertise made no difference in the results.

But when we watch soccer on TV, we usually see replays of the important plays from different angles. Perhaps when several angles are combined, our representation of the visual memory becomes more multidimensional. To test this question, a new group of volunteers was recruited. They were shown two different angles of each play: the sideline view and the view from behind the goal. Then they were tested with still photos, as before. To give you an idea of how this was done, here's a chart showing the different camera angles that were used:

The first and second viewing angles are the positions of the video cameras. Still shots were taken from these spots and three more locations: inbetween the two cameras, before the first camera, and after the second camera. Now the participants were tested with 10 different "correct" photos and 10 distractors, again from a different goal scored by the same team in the same game. Here are those results:

This time, there was no statistical difference between the first and second video angles and the inbetween angle. However, there was a significant difference between these three angles and the "before" and "after" angles. So it appears that seeing more different angles does help us remember other angles, but it doesn't give a perfect three-dimensional representation of the event.

But in film and television, we often see events in a different way: the camera might start with a wide shot, then cut to a different view -- a close up from a different angle. Instead of repeating the entire sequence, the action continues in real time, like the Maradona video I link to above. Some researchers have argued that we place the entire memory in the context of the opening shot. So Garsoffky's team designed a final experiment, where half of the goal-scoring sequence was shown from one angle, then the second half, rotated by 90 degrees, was cut in seamlessly. Participants were shown still shots from both portions of the video, offset at varying degrees from the actual angle they had seen initially. Here are those results:

Now viewers were more accurate only at the identical angle for the portion of the video where the still photos were presented. Watching half the video at a different angle didn't help viewers recognize scenes at any angle, including the other viewing angle they had seen in the video (this is the 90° data point on the chart).

Garsoffky's team argues that this evidence demonstrates that visual memories are primarily two-dimensional: that we remember best from the vantage point where we initially saw a scene. Even seeing a scene from two different angles doesn't improve visual memory much. Perhaps after seeing a scene from many different angles, eventually a complete three-dimensional representation will be built, but even among experts, visual memories are primarily formed in two dimensions.

Garsoffky, B., Schwan, S., & Hesse, F.W. (2002). Viewpoint dependency in the recognition of dynamic scenes. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28(6), 1035-1050.