What Pac-Man tells us about how we recognize shapes

Have you seen this "illusion" before?

The arrangement of the pacman shapes leads you to perceive rectangles, which are actually just empty spaces between the pacmen (that's a technical term -- it's in a journal article, so it must science!). Technically the rectangles are called "Kanizsa-type subjective contours," because while we perceive rectangles, there aren't actually rectangles there. It's a powerful effect, and once you see the shapes, it's difficult to look at the picture without perceiving them. Here's the identical figure with the pacmen rotated in random directions:

See? No rectangles. But the effect is powerful enough that it makes you wonder whether the rectangles are perceived the same way real shapes are perceived. When you look at this figure, for example, you can quickly tell that one of the rectangles is turned sideways compared to the rest of them:

Do we do that as quickly with the pacman rectangles?

A team led by Xingshan Li showed students randomized clusters of rectangles, half of which included a sideways rectangle. Some of the time the rectangles were only suggested using the pacmen, and some of the time actual rectangles were shown, as in the example above. Finally, some of the time both the pacmen and the real rectangles were combined in a single figure, like this:

If we perceive pacman rectangles just like ordinary rectangles, then there should be no difference in the time it takes to determine whether one of the rectangles is sideways. Here are the results:

As the number of rectangles increases, it becomes more difficult to determine if one of the rectangles is different, especially if they're actually all the same. But the difficulty increases significantly faster when the rectangles are only defined by pacmen. This suggests that we may be using a different mental process to recognize and identify pacman rectangles compared to non-pacman rectangles.

So is this true whenever the boundaries of a rectangle aren't defined by a line? In another experiment, Li's team showed people rectangles that were defined in a different way, using vertical lines:

The rectangle on the upper left is really defined in a similar way to the Kanizsa-type Pacman rectangles, by the absence of vertical lines. The other three have borders that vary in their degree of subtlety. The researchers repeated their experiment using these rectangles instead of the pacman rectangles. This time there was no significant difference in response times, regardless of how well-defined the border of the rectangles were.

So we can perceive "shapes" in many different ways, and some of the ways we do it require the use of different cognitive processes. Some of these processes are more efficient than others. To take another example, try to find the letter "L" in these two grids:

It's a lot easier to spot in the grid on the right than on the left, because searching for a color is more automatic than searching for a subtly different shape. Spotting pacman rectangles is more like spotting the black-and-white Ls, while spotting ordinary rectangles -- and those defined by breaks in lines -- is more like spotting the red L.

X. LI, K. R. CAVE, J. M. WOLFE (2008). Kanizsa-type subjective contours do not guide attentional deployment in visual search but line termination contours do Perception & Psychophysics, 70 (3), 477-488 DOI: 10.3758/pp.70.3.477