Method here is a little iffy I gotta say. They had people just watch video and judge which direction the baby was looking, then average the closest two.

also the total difference is like a second and a half out of 70.

also also a parent had to hold the child which introduces a mechanism of biasing.