How do we recognise faces?Most humans can recognise hundreds of faces, and tell the identity of each one, even in different lighting conditions or when someone has changed their hair colour or aged considerably. But how this works, and how our brain codes for individual people’s identity isn’t known. I spoke to Dr Meike Ramon from the University of Glasgow and University of Louvain, who studies face recognition, to uncover why we still don’t understand this seemingly simple ability.
One famous study using single cell recordings from people undergoing brain surgery claimed to discover a ‘Jennifer Aniston Neuron’. The neuroscientists believed they had found a neuron that responded specifically to pictures of the movie star, suggesting information about the identity of a face may be stored in a single cell. However, these findings should be taken with caution. First, an experiment is always limited in terms of the number and type of stimuli it can test for. Second, we don’t have enough individual cells in our brains to assign one to each of the people we have come into contact with, meaning it can’t be as simple as one cell for each person. It was also discovered that the so called ‘Jennifer Aniston cell’ responded to information relating to her, as well as images, suggesting identity is more complex than was first thought.
|Prosopagnosia, or face blindness, is an impairment of the ability to recognise faces that have been seen before.|
Image © Yelisa van der Bij (CC BY)
Since the invention of brain-scanning techniques, researchers have tried to find regions of the brain that are involved in face processing; probably the best studied of these is the Fusiform Face Area (FFA). But while this area is activated more when subjects are looking at faces than at other objects, it is also activated in chess grand masters viewing chess board layouts, suggesting that the area might be engaged when looking at anything we are experts in, which for most of us includes faces. It also shows increased activation towards curved and symmetrical images, leading to the idea that this area is involved in extracting global patterns for discrimination. As important as this region may be, it is now clear that it alone is not sufficient for face recognition, which seems to depend on a network of regions.
One way of studying our normal ability to recognise the faces of people we know is to look at people who don’t have this ability. These people suffer from a condition known as prosopagnosia, or ‘face-blindness’, which can be present from birth or brought on by brain damage. While they may be unaffected in other areas of cognition, prosopagnosia cannot recognise celebrities, their loved ones, or even themselves without using other clues like voice and posture. This is an ability most of us take for granted, but losing it can have a huge impact on a person’s life.
Those patients whose symptoms are brought on by damage to specific areas of the brain offer fascinating insight into face processing - if damage to a region causes prosopagnosia, we can safely assume that that region is vital for face recognition. Interestingly, damage to several different regions of the brain can bring on prosopagnosia, supporting the idea that a complex pattern of connected regions is needed for the ability. Damage one part of the network, and face processing can be affected. As well as areas like the FFA, which have traditionally been linked to face processing, medial temporal lobe regions have been implicated in this pattern. These are areas known to be involved in memory and in judging the importance of objects in our environment - suggesting again that face recognition isn’t a simple process.
Interestingly, the network of areas involved don’t seem to have a simple hierarchical pattern, as is seen in more general visual processing. Normally, when we see an object, information is passed from one brain region to the next to the next, with the complexity of the information increasing at each step. However this is not what happens when we process faces. We know this because patients with damage to one area still show activity in all the others - if it were hierarchical, damage to one region should knock out activity in all those downstream of it. Instead, it seems the areas are connected with some kind of complex feedback loop, communicating with each other on different levels. To fully map this pattern of connectivity is a huge challenge, but one that will greatly increase our understanding of how we process faces, and of visual processing in general.
So if we now know that face processing involves more than just one area, why has the Fusiform Face Area historically been picked out as so important? One answer is that it is easy to identify, and it shows up reliably on brain scans. Other areas now thought to be involved in face processing are in a region of the brain that it is difficult to get a good signal from during scans, because air pockets located nearby can interfere with the signal. This simple artefact of the methods used to look inside the brain may have held back our understanding of face processing.
Familiar vs Unfamiliar faces
One of the big gaps in our understanding is how our brains process faces we are familiar with compared to unfamiliar faces. Most studies have been done using faces that are unknown to the person looking at them, but we quickly become acquainted with individuals and spend most of our lives surrounded by people we do know. It is therefore surprising that so little is understood about how we process familiar faces.
We do know that processing familiar faces is highly efficient, compared to unfamiliar. In one study, participants were asked to sort images into piles - one for each person who was represented. The images were of faces, but taken from different angles and under different lighting conditions. Interestingly, when the images were of people known to them, participants could successfully work out that there were only two different people represented. When the people were unfamiliar, though, participants thought there were an average of 13 different people represented in the photos. This fascinating finding raises the question; how do we store a representations of familiar faces that allows us to disregard lighting, angles etc and see them as them?
If we can work out how the brain disregards unimportant factors to hone in on what it is that distinguishes one person from another, this could lead to important advances in technology. Currently, despite our limitations dealing with unknown faces, humans still outperform computers at facial recognition. Dr Ramon said
While some current algorithms can compete with humans when unfamiliar face matching is involved, face identification—as achieved for familiar faces despite changes of e.g. illumination and pose—remains a serious challenge. Improving the facial recognition abilities of computers could improve security at airports, as automatic passport checking becomes more and more common, and be of benefit in many other applications.
Why are some people better at recognising faces than others?
I’m at a party, or a conference when someone approaches me.
“Hi, Ginny, great to see you again!”
I am filled with panic. Who is this person? Where do I know them from? While often I will recognise their face, I consistently struggle to put a name to that face, or even recall where it was I met them before. I reply enthusiastically, hoping that as the conversation continues, something will jog my memory before it becomes too embarrassing…
This is a scenario that happens to me relatively frequently. It is partly because my job means I talk to multiple people every week, but it is partly something to do with the way my brain works. My mum never has this problem - in fact, she is usually the one on the other side of the conversation, having to remind someone that they met once, 20 years ago, at some event or other. On the other end of the scale, my dad will fail to recognise the main character from a film we are watching if she leaves the screen and returns with a different hair style.
But why do some people have such fantastic memories for faces, while others find it so much harder? This is something scientists are still trying to understand. There is some evidence that there are a group of people, to which my mum probably belongs, known as ‘super-recognisers’. These people are able to recognise faces they know, years after they have last seen them. They are even able to identify celebrities based on pictures of them as children. Somehow, they can remember the essence of someone’s face- that thing that doesn’t change with age or hairstyle, but makes a person look like themselves throughout their lives.
I have asked my mum before how she does this, and she told me that it is all to do with the ratios between their eyes, nose and mouth. While, obviously, my sample of one is not scientific, the ways in which we use these ratios is something that scientists, including Dr Ramon are working on. Hopefully, understanding what is different in the brains of these super-recognisers, and how the representation of a face changes when we become familiar with people could help shed light more generally on how we are able to recognise faces, and even how our visual processing system works as a whole.
This article was written by TWDK's Natural Sciences editor Ginny Smith. Ginny can be found on twitter as @GinnyFBSmith.
why don't all references have links?
 Quiroga, Rodrigo Quian. "Concept cells: the building blocks of declarative memory functions." Nature Reviews Neuroscience 13.8 (2012): 587-597.
 Kanwisher, Nancy, Josh McDermott, and Marvin M Chun. "The fusiform face area: a module in human extrastriate cortex specialized for face perception." The Journal of Neuroscience 17.11 (1997): 4302-4311.
 Righi, Giulia, and Michael J Tarr. "Are chess experts any different from face, bird, or Greeble experts?." Journal of Vision 4.8 (2004): 504-504. doi:10.1167/4.8.504
 Caldara, Roberto et al. "The fusiform face area is tuned for curvilinear patterns with more high-contrasted elements in the upper part." Neuroimage 31.1 (2006): 313-319.
 Rossion, Bruno, Bernard Hanseeuw, and Laurence Dricot. "Defining face perception areas in the human brain: a large-scale factorial fMRI face localizer analysis." Brain and cognition 79.2 (2012): 138-157.
 Bouvier, Seth E, and Stephen A Engel. "Behavioral deficits and cortical damage loci in cerebral achromatopsia." Cerebral Cortex 16.2 (2006): 183-191.
 Barense, Morgan D, Richard NA Henson, and Kim S Graham. "Perception and conception: temporal lobe activity during complex discriminations of familiar and novel faces and objects." Journal of Cognitive Neuroscience 23.10 (2011): 3052-3067.
 Riesenhuber, Maximilian, and Tomaso Poggio. "Hierarchical models of object recognition in cortex." Nature neuroscience 2.11 (1999): 1019-1025.
 Rossion, Bruno. "Understanding face perception by means of prosopagnosia and neuroimaging." Frontiers in Bioscience (Elite Ed.) (2014): 1,258-307.
 Rajimehr, Reza, Jeremy C Young, and Roger BH Tootell. "An anterior temporal face patch in human cortex, predicted by macaque maps." Proceedings of the National Academy of Sciences 106.6 (2009): 1995-2000. DOI: 10.1073/pnas.0807304106
 Andrews, Sally et al. "Telling faces together: Learning new faces through exposure to multiple instances." The Quarterly Journal of Experimental Psychology 68.10 (2015): 2041-2050. DOI:10.1080/17470218.2014.1003949
 O'Toole, Alice J et al. "Face recognition algorithms surpass humans matching faces over changes in illumination." Pattern Analysis and Machine Intelligence, IEEE Transactions on 29.9 (2007): 1642-1646.
 Hancock, Peter JB, Vicki Bruce, and A Mike Burton. "Recognition of unfamiliar faces." Trends in cognitive sciences 4.9 (2000): 330-337.
 Russell, Richard, Brad Duchaine, and Ken Nakayama. "Super-recognizers: People with extraordinary face recognition ability." Psychonomic bulletin & review 16.2 (2009): 252-257.
 Ramon, Meike. "Perception of global facial geometry is modulated through experience." PeerJ 3 (2015): e850.