Faces of Perception

It's tough to explain how people so easily tell one face from another

Newborn babies are wrinkled, wide-eyed strangers in a strange land of light, shadow, and color. Nonetheless, these little bundles of visual innocence take an immediate shine to faces.

Participants in one set of visual-discrimination studies learn to discern differences among imaginary creatures called greebles, shown here in versions designated as male (above) and female (below) by researchers. M. Tarr and I. Gauthier

M. Tarr and I. Gauthier

Taken at the solid line in whole-brain image (left), a computer-generated slice (right) highlights areas, including the fusiform gyrus, that participate in face recognition. M. Tarr and I. Gauthier

Just a few hours after birth, infants begin to imitate adults’ smiles, frowns, and other expressions. Given a choice, the same babies gaze longer at a picture of their mother’s face than at an image of the face of a female stranger. They also boast a budding aptitude for telling strangers’ faces apart and give particular notice to faces rated as attractive by adults.

The magnetic pull that faces exert on babies’ attention has stimulated much research by psychologists over the past 30 years. In related work, neuroscientists are pursuing the brain areas that enable most adults to say with confidence, “I never forget a face.” Debate over the meaning of all this research has now come to a head.

Some scientists suspect that newborns possess an innate ability to spot basic facial features, such as two eyes situated above a mouth. As a result of humanity’s extended evolution in small groups, where it’s critical to discern friends from foes, genes now give human brains a head start in decoding faces, according to these researchers.

Consequently, they add, a few brain areas—in particular, a small patch of right-brain tissue just behind the ear—specialize in perceiving and recognizing faces. People who incur damage to these regions neither recognize the faces of those they know nor remember new faces.

An alternate explanation of face recognition, however, is gaining momentum. Its adherents argue that infants come equipped not with a special face-recognition capability but only with preferences for general perceptual features, such as curved contours. Babies use these visual inclinations as a launching pad for learning to recognize faces, say researchers in this camp.

A propensity for looking at faces may coexist with preferences for several other visual categories in newborns, none of which is predetermined by genes, these researchers argue. By 3 to 4 months of age, infants usually prefer pictures of cats over horses, tigers over cats, chairs over tables, and mammals over birds. Such curious predilections probably exist even in newborns, these scientists propose.

Moreover, they point out, the behind-the-ear brain area most directly implicated in face recognition actually coordinates all sorts of expert visual judgments. It’s just as crucial for making deft distinctions among classic cars, bird species, and imaginary creatures with no faces as it is for recognizing pictures of one’s high school classmates. Damage to this part of the brain hinders object recognition to a lesser extent than it does face recognition, but the effect is still noticeable, add supporters of this view.

The theoretical division runs deep among face researchers. “Debate about the nature of face recognition is unbelievably heated right now,” says neuroscientist Charles A. Nelson of the University of Minnesota in Minneapolis. “It has polarized groups of researchers.”

In the long run, this dispute will yield major scientific insights, holds neuroscientist Nancy Kanwisher of the Massachusetts Institute of Technology. “It will take a while to figure out how the brain carves up visual perception,” she says. “This is how science works.”

A fuzzy world

It’s hard to know precisely what a baby sees when he or she looks at a face looming overhead. Studies indicate that newborns perceive a fuzzy world, devoid of sharp delineations among objects and the nuances of noses, cheeks, and other parts of the facial landscape.

Despite their vague take on the world, infants have an innate preference for gazing at facelike sights, such as a pair of round blobs over a horizontal line, contends psychologist Mark H. Johnson of Birkbeck College in London.

“Newborns possess a simple, blurred representation of facial structure,” Johnson says.

His view derives from several findings. First, newborns see well enough to imitate adults’ facial expressions. Second, babies so quickly start to distinguish among faces and recognize their mother’s face that there is virtually no time for learning these tricks. Third, within days of birth, infants know enough about facial features that they prefer attractive over unattractive faces.

Finally, as with adults, babies show no partiality to good-looking faces if the images of the faces are turned upside down.

To do all this, babies must begin life knowing how the human face is organized, contends psychologist Alan Slater of the University of Exeter, England. Newborns thus naturally look toward facelike configurations, with two eyes above a nose and mouth, he theorizes.

Other scientists see more merit in visual simplicity. Newborns probably focus on general elements of faces, contends psychologist Paul C. Quinn of Washington and Jefferson College in Washington, Pa. The contour, or shape, of the head may represent a particularly compelling component of the face for babies, he says.

For instance, in the January Journal of Experimental Child Psychology, Quinn and his coworkers reported that 3- and 4-month-old infants distinguish among cats and dogs based on silhouettes of the animals’ heads.

Consider babies who first spent time looking at a cat’s head silhouette. In a later trial, these infants looked longer at a head silhouette of either a new cat or a dog than at the silhouette of the previously seen cat, revealing a preference for visual novelty. Along the same line, babies who viewed a series of head silhouettes of either cats or dogs later preferred to look at head silhouettes of the other species.

In contrast, no such signs of appreciating novel figures appeared among infants shown body silhouettes of either cats or dogs.

In earlier work, Quinn had found that 3- and 4-month-olds also distinguish among familiar and novel silhouettes of human heads. At that age, babies favor head shapes of people over those of cats, dogs, and other animals, he says.

Babies soak up the subtleties of the faces they encounter with surprising ease, Quinn adds. In one stark example, he finds that 7-month-old infants of both sexes raised by single mothers prefer to look at pictures of female faces, whereas those raised by single fathers opt for male faces.

Face recognition

In fact, constant exposure to faces in the first few years of life generates the brain’s face-recognition system, argues Nelson in the March-June Infant and Child Development. He suggests that babies only need to start out with a capacity for discerning general properties of figures, much as Quinn proposes.

Experience in looking at faces during the first year of life focuses infants’ perceptual spotlight, Nelson contends. For instance, he and his coworkers find that infants do better than adults at noticing facial differences in monkeys and also in people from racial backgrounds other than their own. These advantages illuminate babies’ reliance on broad visual cues that, in Nelson’s opinion, eventually get replaced by a system for rapidly recognizing familiar human faces.

Faces contain two structural elements that are big-time baby pleasers but are by no means unique to faces, according to studies directed by psychologist Francesca Simion of the University of Padova, in Italy. First, newborns prefer to look at patterns with a greater number of elements in the top half of the visual field. They gaze longer at a T-shape pattern composed of squares than at an upside-down T shape, as well as at head shapes with more squares randomly positioned in the top than in the bottom half.

Second, in the days after birth, babies fixate more readily on curved rather than straight contours. Infants shown face drawings without head contours exhibit none of their usual preference for an upright face, Simion’s team finds.

Babies aren’t the only ones who can use such basic visual cues to grasp faces. A computer network programmed to respond to simple visual features and to modify its output after getting feedback on errors learns to recognize faces with considerable accuracy. This system, devised by Hervé Abdi of the University of Texas at Dallas and his colleagues, captures increasingly fine visual details as it encounters a greater number of faces.

“The bulk of the evidence suggests that the ability to recognize faces is learned,” Nelson argues.

Special status

Controversy also surrounds efforts, conducted mainly with adults, to track down brain areas devoted to face recognition.

Evidence that faces enjoy special status in the brain comes from individuals with a condition called prosopagnosia. After suffering damage to the brain’s right temporal lobe, these people have lost the ability to recognize familiar faces. Moreover, a different pattern of temporal lobe brain damage blocks the capability to identify inanimate objects, while sparing face recognition.

A 16-year-old boy named Adam presents a particularly intriguing case of prosopagnosia. Adam’s temporal-brain damage occurred when he contracted meningitis at 1 day of age. He now exhibits “impressively bad” face recognition and mild difficulty at detecting various objects, says psychologist Martha J. Farah of the University of Pennsylvania in Philadelphia. Adam draws a blank even for pictures of his mother.

Uninjured parts of Adam’s young brain apparently had no way to regroup and assume responsibility for face recognition, Farah proposes. “This indicates that the genome encodes for a specific face-processing area in the brain that can’t be replaced,” she says.

Several brain-imaging studies suggest that face recognition depends on a few regions of the temporal lobes, particularly the area just behind the right ear known as the fusiform gyrus. Kanwisher and other investigators find that neural activity in the fusiform gyrus surges at least twice as strongly when healthy adults view faces as when they look at assorted objects, letter strings, or the backs of human heads.

The findings from brain-damaged patients and healthy volunteers justify renaming the fusiform gyrus as “the fusiform face area,” Kanwisher says.

You could just as easily dub this patch of tissue “the fusiform greeble area,” according to Isabel Gauthier of Vanderbilt University in Nashville, who has directed a series of brain-imaging studies on object recognition. After a little training, most volunteers easily classify imaginary, faceless, plantlike creatures called greebles according to their sex and family. These participants display elevated fusiform gyrus activity as they identify pairs of matching greebles.

Comparable brain responses occur as people peruse real-world objects or animals that they know much about, Gauthier and her coworkers find. For instance, the fusiform gyrus lights up in car buffs as they examine pictures of different makes and models of classic automobiles and in bird authorities as they inspect images of various avian species (SN: 2/5/00, p. 91).

Rather than specializing solely in faces, the fusiform gyrus fosters proficiency at visually analyzing any class of items that a person strives to master, Gauthier asserts. “The brain doesn’t need to care about different categories of things in the world, such as faces,” she remarks. Instead, neural tissue solves computational problems, such as assembling visual elements into meaningful wholes.

Kanwisher remains skeptical of Gauthier’s findings until the data on greebles, cars, and birds are confirmed by independent researchers. Gauthier, who did postdoctoral work in Kanwisher’s lab, responds that she looks forward to such attempts.

It’s unclear whether infants and children learn to identify faces in the same way that adults gain visual expertise with whatever strikes their fancy, Gauthier adds. In a recent study, she and her colleagues tested young adults with autism, a disorder that includes difficulty with face recognition but involves no apparent temporal lobe damage. Nevertheless, the participants exhibited unusually low fusiform gyrus activity while trying to tell faces apart.

Whatever mechanisms enable neurologically healthy kids to develop face expertise in the fusiform gyrus may be unavailable to their peers with autism, Gauthier theorizes.

She now wants a closer examination of brain-damaged patients, such as Adam. Having been told that their injury should cause problems with recognizing faces, prosopagnosics usually give up quickly on tests of this ability, she says. In contrast, the same individuals assume they should routinely conquer object-recognition tasks, so they spend a lot of time poring over such tests. As a result, the relative superiority of object recognition among prosopagnosics has been exaggerated, in Gauthier’s view.

Investigators have yet to assess whether Adam looks longer at objects than at faces on recognition tests, she notes.

Temporal lob activity

Intriguing leads from other research groups seem unlikely to break the theoretical stalemate over face recognition.

Distinct, overlapping patterns of temporal lobe activity occur when volunteers look at faces, chairs, shoes, cats, and other remarkably specific categories of objects, according to brain-imaging studies directed by James V. Haxby of the National Institute of Mental Health in Bethesda, Md.

“We can tell if a person is looking at a shoe, a chair, or a face, based on the pattern of their brain activity,” Haxby remarks. He described such findings in March at the annual meeting of the Cognitive Neuroscience Society in New York City.

Specific neural-activity patterns serve as maps of the visual features that characterize each category, Haxby proposes. Overlapping maps occur because categories often share some visual elements.

It’s unclear how the brain generates maps for different categories. There’s clearly not a genetic capacity for distinguishing chairs from shoes, although one might exist for telling faces apart, says Haxby.

Whatever the case, the first few months of life appear to be critical for forming the neural framework of face recognition. A team of McMaster University psychologists in Hamilton, Ont., examined 14 young people, 9 to 21 years old, who were born with cataracts in both eyes and were blind until they had eye surgery at 2 to 6 months of age. They now have great difficulty in seeing faces as unified entities, the team reports in the April 19 Nature.

When shown a pair of face pictures that differ only in the configuration of the same features—say, one face with closely set eyes and little space between the nose and upper lip and another face with eyes wide apart and more space above the upper lip—the former cataract patients usually couldn’t distinguish one picture from the other. Volunteers of the same age but with no previous vision problems did far better at this task.

Further testing of these people suggests that adults who were deprived of vision early in life manage to recognize faces by focusing on the shape of the nose and other isolated features, says study coauthor Richard Le Grand.

Early visual input to the right brain, which arrives via the left eye, proves vital for perceiving faces as integrated units, Le Grand adds. People born only with left-eye cataracts that were later removed do just as poorly on face-configuration tasks as those who needed two cataracts removed, he and his coworkers find.

LeGrand’s team plans to use brain-scan technology to identify right-brain areas activated in these individuals during face recognition. The researchers will also see if early-cataract patients can learn to make expert visual distinctions of objects other than faces, such as Gauthier’s greebles.

If such studies don’t change the face of the face-recognition debate, they’ll certainly feed into its body of hotly contested data.


Correction: “Faces of perception” should have credited psychologist Michael J. Tarr of Brown University in Providence, R.I., as codirector of the research in which people learned to discern differences among imaginary creatures called greebles.

Bruce Bower has written about the behavioral sciences for Science News since 1984. He writes about psychology, anthropology, archaeology and mental health issues.

From the Nature Index

Paid Content