You probably don't even question whether you'd notice a major visual change in your environment. If the person behind the lunch counter
were suddenly someone else, or the family dog changed color, you'd notice, right?
Not necessarily. Though you may not know it, because the action is so momentary, your eyes regularly dart to the side -- a movement called a saccade. And if something, even something major, changes while you're saccading, you just won't notice it.
In an experimental setting, scientists can determine when you're saccading. They'll have you look at a picture of, say, a boy with an enormous parrot on his shoulder (one of the images used by Grimes, 1996). Then, as you saccade, the color of the parrot will change from green to red... and you won't notice. And this is bright green and bright red we're talking about here, and a parrot that takes up roughly 25% of the picture area.
You don't have to be saccading for this to work -- an artificial interruption will do the trick. Simons and Levin (1998) had an experimenter go up to random people on the street and ask for change for a dollar. While the subject fumbled for coins, two people carrying a door (!) would walk between the subject and the experimenter. When the door was gone, there would be a different experimenter there to accept the change... but only half the subjects noticed, even though Experimenter #2 had different clothes, build, haircut, and voice from Experimenter #1.
Simons, Reimer, and Franconeri (2000) showed that, in fact, you don't even really need the artificial interruption: If the change occurs slowly enough, subjects still won't notice. Their subjects looked at a scene, knowing something was going to change, actively searching for changes... but even when the change was a large building slowly fading out of a street scene, people just could not tell.
One explanation of change blindness, from a Harvard researcher:
If observers could "take in" an entire scene with a single attentional fixation, they could detect changes anywhere in an image with equal facility. Instead, observers apparently must scan an image, encoding the scene piecemeal (Rensink et al., 1997). In order to retain information about an object or its properties from one view to the next, observers must re-code the information, explicitly comparing the abstracted representation of the initial object to the changed object (Simons, 1996). Objects that are not re-coded are not remembered in any detail. Given the number of potential features and objects in a typical natural scene (effectively an infinite number), many, if not most aspects of a scene will not be preserved across views. Because observers are more likely to focus attention on important objects, they are more likely to notice changes to objects in the center of interest of a scene.
It's pretty amazing to me that our level of natural obliviousness is this high... or, at least, that this can be empirically proven.
and an excited philosophy/cog. sci. Ph.D.