I finally had a chance to try the Beatles Rock Band video game. It’s terrific, exhilarating fun. And it’s just a bit creepy. Its creepiness stems from the game’s depiction of the Beatles themselves. The faces of John, Paul, George and Ringo are completely recognizable - as well as dynamic, and emotive. But they also have a characteristic common to most computer-animated faces designed to look like recognizable people. The faces are just familiar and real enough to notice that they’re unreal. And this makes them look a little eerie.
This is a well-known phenomenon in computer animation and robotics called the “Uncanny Valley”. The concept was originally discussed by robot designer Masahiro Mori to describe how we react to seeing humanoid forms—robots, animated characters—as their appearance ranges from less to more human. Up to a point, we find human-like characteristics appealing and even endearing in robots and animated characters (think C-3PO and Bart Simpson). But when their appearance—particularly that of their face—becomes almost life-like, we often find them eerie, and zombie-like. Their appearance falls into a valley where it seems uncanny.
You’ve likely noticed an eeriness to some computer-animated faces in films, TV, or videogames. The characters in the 2004 animated film “The Polar Express” are well-known examples. At the time of its release, reviewers described these faces as “unnerving”, “creepily unlifelike”, and possessing a “chilly, zombie-like aura”. Other examples of uncanny valley faces include those seen in the films “Final Fantasy” and “Beuwolf”; the face of Tiger Woods in the eponymous video game (currently, the least of Tiger’s problems); and that of the long deceased Orville Redenbacher who was revived for a 2004 popcorn commercial. All of these faces were rendered to look as realistically human as possible. Still, their slight imperfections made them seem eerily not-quite human.
Why do we have such visceral reactions to quasi-realistic human appearances? One theory is that our queasiness is a by-product of evolutionary pressures to avoid corpses, and their possible diseases. But why does the uncanny valley effect occur for human faces in particular? We don’t, after all, react so negatively to imperfections in animated dog, cat or giant ape faces. The answer likely lies with our supreme perceptual sensitivity to human faces. Decades of laboratory research has shown that we are exceptionally sensitive to (upright) faces, finding them impossible to ignore, and easily making very subtle distinctions. (We notice when the distance between a face’s eyes changes as little as one-tenth of an inch). Faces residing in the uncanny valley click in our brain’s super-sensitivity, allowing us to become aware of even the subtlest deviations from real.
Research on the uncanny valley has shown that the more real a computer animated face is rendered –using for example, photorealistic skin– the less tolerant we are of small deviations in facial shape, as well as feature size and location. This fact is particularly true of deviations in the size, position, and details of the eyes – one of the eeriest features of The Polar Express characters. Similar effects have been found for our tolerance of facial movements: the more photorealistic the face, the less tolerant we become of unnatural movements. In fact, a general ‘all or none’ principle is emerging from this research: if any characteristic of a face is rendered in a convincingly realistic way, all other characteristics must be realistic or the face can seem eerie.
This is something Steve Preeg and his colleagues at Digital Domain (a Los Angeles based film post-production company) know all too well. They were responsible for creating what many consider the first convincingly realistic computer-animated face: That of Brad Pitt in the 2008 film, The Curious Case of Benjamin Button. If you’ve seen the film, you may have had a sense that computer graphics were used to create the effect of an old Brad Pitt face on a small body. But, you probably weren’t aware that for the first 50 minutes of the film, you never actually saw Pitt’s real face. Instead you saw a purely computer animated face –a face so realistic, that it fooled other animators. Benjamin Button is largely considered the first animated character to have traversed the uncanny valley.
Steve Preeg and his group were successful, and won an Academy Award, because of their adoption of new technologies and their extraordinary attention to detail —including the all-or-none constraint of the uncanny valley. According to Preeg: “It was a constant process of fixing one problem, and then noticing something else. Someone would notice that the lighting on Benjamin’s face didn’t quite match the background. We’d go back, fix this, and then notice that the skin didn’t look quite right when he turned his head. The uncanny valley is an evil, evil place.”
In fairness to the Beatles Rock Band, the game’s creators certainly didn’t have the budget and resources afforded Preeg’s group. Instead, the game’s animators reportedly rendered the Beatles’ faces a bit cartoon-ish so as to not draw attention to their quasi-human likenesses. Still, our hyper-picky brains can’t help but notice when animated faces aren’t quite right, allowing very few characters to cross what Steve Preeg calls “the uncanny valley of death”.
Lawrence Rosenblum is a Professor of Psychology at the University of California, Riverside. He studies multimodal speech perception and general auditory perception. His book on our implicit perceptual skills, “See What I’m Saying: The Extraordinary Powers of Our Five Senses” (www.lawrencerosenblum.com) will be published by Norton Press in March.
Ge, L., Luo, J., Nishimura, M., & Lee, J. (2003). The lasting impression of Chairman Mao: Hyperfidelity of familiar-face memory. Perception, 32, 601–614.
Green, R. D., MacDorman, K. F., Ho, C.-C., & Vasudevan, S. K. (2008). Sensitivity to the proportions of faces that vary in human likeness. Computers in Human Behavior, 24, 2456–2474.
Mori, M. (1970). The uncanny valley. Energy, 7(4), 33–35.
Young, A. W. (1998). Face & mind. Oxford: Oxford University Press.