The notion of the "Uncanny Valley", where artificial people are more disturbing the closer they come to "reality", has been around for ages. But now, thanks to a research team in San Diego, it's bonafide science.
As reported on Wired, The University of California confronted twenty test subjects with a variety of videos showing a very special kind of robot performing basic actions. That robot is the Repliee Q2, from Osaka University, which is based on an actual person and is capable of a wide range of intricate facial expressions.
Three sets of footage were compiled: one of Repliee "naked", her wiring and internal components exposed, one of the robot with her "human face" on and a third of the actual human being the robot is based on performing the same functions.
While the subjects were shown the footage, they were wired to an fMRI machine, which showed that while things were fine during the human and naked robot demonstrations, when they watched Repliee with her artificial face on, their brains freaked out.
It seems the issue was with our brains not liking the disparity between superficial human appearance and robotic actions. It expects human things to act human, and when they don't, it throws up a flag.
As the Uncanny Valley is such a common problem in video games, the team's findings make you wonder whether developers are going about overcoming it the right way. If it's all in the motion maybe they need to worry more about the way a game character moves (in particular their faces) instead of how well they're textured.