Researchers found that “super-recognizers” usually need only one or two features.
From
To