Criminal Justice Degree: Guide to Face Recognition

How do we recognize faces?  Is there some specific feature we look for?  Or is it a very specific combination of shapes and colors?  Is there one mode of thought that unites the high technology of computer face recognition and the very, very low technology of a baby recognizing its mother’s face?  In this students’ guide, we break down the basics of this complex, interdisciplinary area of study– a subject that covers topics as diverse as philosophy of mind, neurology, psychoanalysis, and computer science.

Face Recognition: Nature versus Nurture

At birth, babies cannot differentiate between faces; this much is known.  By three months, babies can generally recognize faces.  Some scientists suggest that babies have an innate ability to recognize faces, while others suggest that the ability to learn what faces look like is part of a broader ability to recognize patterns.  As far as physical, statistical similarities go, human faces are very similar.  However, we think of each face as unique and highly differentiated.  This suggests a strong ability to recognize subtle differences.  However, whether this capability is learned or acquired is unclear.  Specific areas of the brain have been shown to correspond to facial identification, but when newborns see faces, these areas aren’t activated.

Facial Recognition Among Adults

Psychologists and neuroscientists have offered a number of competing theories regarding the mechanisms by which people recognize faces.  We don’t have the space to discuss most of them, but we’ll cover one of the most popular.  It’s also the basis for many other face-perception theories.

Like many theories, it suggests that face recognition is the cumulative result of a number of independent processes, which may have either evolved independently or may have simply been socially communicated over the generations.  In this model, memory plays a critical role, and the mind makes analogies between faces it sees and faces it remembers, using metrics such as age, skin color, distinctive features, and more subjective measures such as attractiveness.  And, whether or not we’re aware of it, we split the face into a number of features, which enables us to recognize faces from multiple angles.  This semantic memory is also used to recognize emotion in a variety of different faces, and to derive information about age, gender, race, etc. from individual faces.

Some of the most fascinating research in this area comes from the study of prosopagnosia, a condition widely referred to as “face blindness.”  Face-blind individuals have difficulty making these recognitions.  Some of the most extreme cases, victims of a subset of the condition known as apperceptive prosopagnosia, are unable to distinguish between two photos of two different faces.  The use of other identification techniques, such as hair, clothing, and voice often allows face-blind people to get by; many are not even aware that they are face-blind.  Interestingly, two of the most prominent researchers studying prosopagnosia– Ken Nakayama and Oliver Sacks– are face-blind themselves, as is Chuck Close, a highly regarded contemporary artist best known for his blown-up images of human faces.  A test known as the Glasgow Face Matching Test is widely used to study prosopagnosia, as well as in other studies on face recognition.

Face and Race

Humans have a tendency to perceive facial differences within their own ethnic group better than outside it.  This has often led to the perception that people of other races “all look the same.”  Many researchers have suggested that this effect– often referred to as the cross-race effect– is stronger among some cultures than others.  However, these results are heavily disputed.  Some studies show that white Americans are more likely to only recognize differences within their own race, whereas other studies show that black Americans are more likely.  Cultural values have been suggested to play a major role in how developed a cross-race effect is.  Individual experience also informs the effect.  For instance, people who are raised primarily among people of other races are more likely to recognize facial differences outside their own race.

Facial Perception and Law

Eyewitness testimonies, especially when it comes to facial recognition, are notoriously unreliable.  People tend to focus on specific features when they remember the face of a criminal, and it’s often something that can be easily modified, such as facial hair or an article of clothing.  Memory of faces is also highly susceptible to modification and suggestion.

The cross-race effect is often compounded with this unreliability.  An eyewitness will probably be able to identify the race of a criminal.  However, if that criminal is of a different race than the witness, the witness’ ability to select his or her face from a police lineup is even more limited than it normally would be.  When compounded with the racist tendencies implicit in American society, this has often led to erroneous accusations of guilt.

Computer Facial Recognition

A certain number of objective metrics can be used to determine human facial structure, and these have been used to inform computer face recognition systems.  Nose size, cheekbone measurement, and eye shape are often used.  Increasingly, three-dimensional models of facial recognition are being employed, which are significantly more accurate.  A third way of improving accuracy is the detection of skin texture, an important measurement that helps to distinguish faces.  Even in less sophisticated programs, techniques have been developed for a computer to identify a human face in a photo as distinct from its surroundings.

Further Reading

We’ve linked to a variety of resources, including scholarly journals, educational sites, and companies working on facial recognition software.

• Face Research is a British educational portal containing in-depth information on facial perception and fun interactive demos.
• How are your facial recognition skills?  Test yourself online.
• The Perceptual Expertise Network is a group of researchers based at Vanderbilt University conducting cognitive studies involving perceptual mechanisms of various kinds, including facial perception.
• The Facelab at the University of Western Australia is another major laboratory conducting facial perception research.
• Global Emotion is a German-based company that develops technologies to identify and counter the cross-race effect.
• Face-rec.org documents the fast-evolving world of computer-based facial recognition.
• This PDF report, hosted by the University of Texas at El Paso, documents the cross-race effect.
• The Pittsburgh Post-Gazette examines the relationship between memory, eyewitness testimony, and race in this 2010 article.
• Wired discusses prosopagnosia.
• Luxand is a company specializing in face recognition technologies.
• HowStuffWorks provides a consumer/non-expert’s introduction to the science of facial recognition by computers.

IMAGE: Computer dentification of faces compared to their surroundings is a major technological application of the theories behind facial perception. (Source: Wikimedia Commons)