people can identify up to 10,000 faces, scientists have discovered. In the first study of its kind, researchers quantified the number of faces an individual human can know—and findings showed people were able to recognize between 1,000 and 10,000, with the average being 5,000 faces.
The research, published in Proceedings of the Royal Society B: Biological Sciences, was carried out by Rob Jenkins, from the Department of Psychology at the University of York, and colleagues.
"The vocabulary of facial identities is large given the context of our species history,” he told Newsweek. “For most of that time, humans lived in groups of around a hundred people. These days, humans may live in cities of millions, and we’re bombarded with faces from the media too.
"The demands on human face recognition ability have intensified dramatically. But it seems that whatever mental apparatus equipped us to differentiate dozens of people also equips us to differentiate thousands of people. That was unexpected.”
For the study, the team asked participants to write down as many faces from their personal lives as they could within an hour. They then did the same for famous people. Researchers found the rate at which people would write down faces started fast then slowed down as they struggled to think of more. The team then worked out the pace at which people slowed down to estimate at what point they would run out of faces altogether.
They liken the number of faces a person knows to individual vocabulary range. This accounts for the huge variation in scores. Jenkins also said he expects they have not found the upper limit in individual performance.
“Our current record holders are up at 10,000, but I expect that record to fall,” he said. “Suppose we had no idea how far humans can run—we could get a good estimate by testing a group of undergraduate students. But it’s unlikely that a random sample would contain the far extremes of human performance.
“The ability to distinguish different individuals is clearly important, as it allows you to keep track of people’s behaviour over time, and to modify your own behaviour accordingly. Given the social lives of our ancestors, the ability to recognise thousands of individuals might seem like overkill. But there are plenty of examples of overkill in nature. The venom of some spiders can kill a horse, even though the spider presumably has no ambitions to eat the horse.”
The findings open up a wide range of research avenues—particularly those concerning facial recognition technology. Jenkins said psychological research shows big differences between faces we do know, and those we don’t. We can reliably identify familiar faces—but how we do this is not known. “A better understanding of familiar face recognition in humans should inform better modelling of that process in machines,” he said.
Megan Papesh, Associate Professor from the Department of Psychology at the Louisiana State University, whose research looks at facial recognition, said the findings were exciting. “People can intuitively appreciate being expert face processors,” she told Newsweek.
“We see faces in clouds, lattes, potato chips, and myriad other objects. This tendency to impose structure where there is none is called pareidolia and shows the amusing power of our familiarity with face patterns. Although researchers have long appreciated that people have the incredible memory for personally familiar faces, Jenkins [and the study authors] provide the first-ever estimate of a face identity ‘vocabulary.’ Personally, I find this research incredibly cool and exciting.”
She said that compared to words, faces are harder to recognize—they are all made up of the same general properties. Our eyes, noses, mouths etc. are all in the same place.
“Despite the relative homogeneity of faces, Jenkins and colleagues found that people know approximately 5,000 identities, which opens the door to many new and interesting research questions. For example, it will now be possible to establish developmental trajectories of face identity learning and to investigate the impact of neurodevelopmental or neurodegenerative disorders that affect social processes. There are many directions this research could go, which I think is the mark of a great study.”