PITTSBURGH _ Recognizing a familiar face accurately and almost instantly _ despite the frown or smile, the new wrinkles, or beard, or added hat or glasses _ involves a mysterious brain process, especially given the many thousands of highly similar faces in a world of billions.
A Carnegie Mellon University study in the Proceedings of the National Academy of Sciences takes important steps in describing the mystery of face recognition, one of the brain's most impressive accomplishments.
"That we are able to recognize tens of thousands of faces with accuracy and rapidity continues to be a scientific mystery," said study leader Marlene Behrmann, who holds an endowed professorship in cognitive neuroscience at CMU. "The question is how this is done, and this study is an attempt to understand that question.
"If we can answer it and lay bare how the brain achieves this, it also offers the possibility that other visual recognition processes may follow the same pattern, including reading, with different type and handwriting."
The study involved CMU's Center for the Neural Basis of Cognition and the University of Toronto Scarborough.
Cascading brain signals confirming identity "feels effortless," said Behrmann, who holds a Ph.D. in cognitive neuroscience. "It's not like this is hard work. No one is sweating over it. But the speed and lack of effort belies the complexity of the brain computations taking place."
What seems "almost magical," she said, actually involves two complicated brain processes.
The first is appearance-based or image-based, with the brain making literal observations of the face including the shape of nose, mouth or eyes, size of cheeks or type of eyebrows.
Next the brain uses those image-based details in a more abstract, "identity-based" process with neural signals traveling throughout the brain in a give-and-take, back-and-forth process, to analyze and link those facial characteristics to a particular person in memory.
In essence, the brain pares down details of the face to key geometric characteristics while using ever more details to match those characteristics to a known face, ultimately having enough evidence to confirm the identify of a friend, family member, long-lost acquaintance, Barack Obama or Donald Trump.
For example, Behrmann said, the brain might focus on a pronounced dimple in the chin, then the large size and shape of the head and consider that the person just might be Jay Leno, only to have further evidence of eyebrows, facial expressions and hair color to confirm the identify.
In an even more complex identity-based procedure, the brain analyzes the facial structure despite variations, different angles, lighting and shading, facial alterations and even the effects of aging if the person hasn't been seen for years.
Recognition typically occurs within a range of a few milliseconds to less than half a second.
In the study, subjects wore helmets containing neuroimaging technology to measure brain signals on the scalp, with simultaneous magnetic resonance imaging or MRI to determine precisely where those specific signals were originating.
Each person was asked to identify 91 similar faces, each face shown twice, once with a smile and once with a neutral expression. Each version of the faces was then seen about 100 times by each participant in the study, for a total of 10,000 trials.
The careful measurement helped the research team to understand how the brain recognizes faces despite variations. The large amounts of data were computer-analyzed to show, millisecond by millisecond, how complicated cycles of brain signaling occur, ultimately resulting in a distinct brain signature for each face. These signatures allow the observer to recognize the face.
"These results have important implications for understanding the rapid emergence of fine-grained, high-level representations of object identity," the study said, describing that process as "a computation essential to human visual expertise."
