March 9, 2004

Facing the facts about recognizing faces

McMaster researchers study how our brains picture faces

Notice anything different about these two pictures? It’s the same person, with one small difference. One image has been altered, but most people won’t see how until they’re viewed upright. Want to see? Click on image to view.

Hamilton, ON - Notice anything different about these two pictures? It’s the same person, with one small difference. One image has been altered, but most people won’t see how until they’re viewed upright.

This is one example of the so-called the “inversion effect” – it’s harder for the brain to process upside-down objects than upright objects, and the inversion effect is especially strong for the perception of faces.

“For most people, it’s easy to recognize a range of faces, even under various lighting conditions and from different views. But when those faces are turned upside-down, we experience problems,” says Allison Sekuler, professor of psychology and Canada Research Chair in Cognitive Neuroscience at McMaster University.

Sekuler says human faces consist of two eyes, a nose, and a mouth, organized in just about the same way for every face. For decades, people thought the face inversion effect meant that the brain uses the information in faces in very different ways to recognize upright and upside-down faces.

Traditionally, recognition of upright faces was thought to hinge on the organization of features across the whole face, whereas recognition of upside-down faces relied much more on identifying local features.

Sekuler and her team set out to test that idea directly. Their results, which will appear in the journal Current Biology on Tuesday, provide an entirely new picture of what goes on when our brains picture faces. To obtain a clear view of how the brain processes information about faces, the researchers actually added “visual noise” (resembling snow on a de-tuned television) to face images. By keeping track of how that “noise” affected perception, the researchers were able to tell what parts of the faces were most important for recognition. Surprisingly, all observers relied mostly on the region around the eyes and eyebrows, regardless of whether the faces were upright or upside-down.

“The devil is in the details,” says Sekuler. “Although most of the relevant information for recognizing our faces was right around the eyes, people seem much more efficient at picking up that information in just the right way when the face is right side-up.” These results fly in the face of previous theories of face recognition. Instead, the researchers suggest that the face inversion effect may be an example of the old saying, “practice makes perfect” – people simply have a lot more experience recognizing upright faces, and that makes them better.

According to this view, the inversion effect is a fascinating example of how the human brain processes information, and how our brains can be trained to process difficult tasks more efficiently. In a related study, to be published in April in the journal Cognitive Science, Sekuler and her research team applied similar “noise” obstructions to faces and unfamiliar textures to determine how people’s recognition skills improved with learning. With both types of patterns, everyone who was tested improved. For faces, people became more efficient at picking out the relevant information around the eyes and eyebrows. For textures, different individuals adopted different strategies for improvement. Although everyone became more efficient at picking out the right details, the locations of those details differed dramatically (some people relied more on information in a top corner, whereas others relied on information in the middle or bottom).

“In working with textures, we found that people learned to recognize them in different ways, even though they all ended up performing the task equally well,” says Sekuler. “For the first time, we were able to get a direct view of what strategies the brain used to improve recognition. Understanding the unconscious learning strategies people use, and how those strategies vary across individuals, will help us to establish more effective training techniques.”

Sekuler hopes that by identifying how the brain normally processes this kind of information, she and her group will be able to develop training programs for people who have impaired facial recognition skills, such as autistic individuals and some stroke victims.

“The first step toward improving performance in impaired populations is to understand how the typical brain processes information,” she says. “With this work, we’ve made a big leap toward that end.”

Sekuler’s research team includes Patrick Bennett, professor of psychology and Canada Research Chair in Vision Science, and Carl Gaspar, graduate student, from McMaster University, and Jason Gold assistant professor of psychology from Indiana University. The work was funded by the Natural Sciences and Engineering Research Council of Canada and the Canada Research Chairs.