Mechanisms of Recognition: Change-Blindness and the Perception of Pictures

...

Just a year later, at a conference on perception in Vancouver, British Columbia, John Grimes of the University of Illinois caused a stir when he described how people shown computer-generated pictures of natural scenes were blind to changes that were made during an eye movement. Dennett was delighted. "I wish in retrospect that I'd been more daring, since the effects are stronger than I claimed," he says.[3]

Grimes’ experiments tested whether subjects noticed changes in a scene introduced between saccades. In his 1996 paper, “On the Failure to Detect Changes in Scenes across Saccades” Grimes concludes that very little information is carried across from saccade to saccade.[4] Grimes’ work prompted a renewed interest in the human ability to detect change and has inspired much speculation and experimentation on the change blindness phenomenon. It raises the question “Is the visual world a grand illusion?”[5].

In Dennett’s view, the whole phenomenology of perception is largely an illusion. The common conception is that there is a seamless stream of consciousness; Dennett’s view is that consciousness is a fragmented, disjointed thing and that we piece it all together in retrospect. Vision researchers who set out to explain how we arrive at the seamless detailed, integrated, deep and colourful reality which we experience are actually falling at the first hurdle. Dennett maintains that the little we do see is largely in black and white, patchy and superficial. That is, we are not registering the world in anything like the detail that we think we are and we do not recall from moment to moment most of what presents to our retinas. Dennett’s experiment with the playing card demonstrates this nicely. He comments:

The visual field seems to naive reflection to be uniformly detailed and focused from the center out to the boundaries, but a simple experiment shows that this is not so. Take a deck of playing cards and remove a card face down, so that you do not yet know which it is. Hold it out at the left or right periphery of your visual field and turn its face to you, being careful to keep looking straight ahead (pick a target spot and keep looking at it). You will find that you cannot tell even if it is red or black or a face card.[6]

You will be aware of motion at the periphery of your visual field but will not be able to identify the card until it is almost right in front of your eyes. The reason that we cannot see colour or identify objects outside 5 degrees of dead centre is because only the receptors (cones) of the fovea around the centre of the retina generate a colour signal. The rest of the receptors (rods) cannot generate a colour signal. These facts are not remarkable – the remarkable thing is that we don’t notice this staggering deficiency in our vision. Given these, and other facts about the limitations of the human visual system (we have two upside down distorted retinal images and our eyes constantly dart around in saccade movements during which the retinal image “greys out”) it is also remarkable that vision research didn’t predict change blindness and inattention blindness earlier.

In fact, the results of the hundreds of change-blindness experiments that have been conducted since 1991 are so surprising that they have generated a new interest in the role of attention in both perception and consciousness. They have also given birth to a new discipline – visual metacognition – the study of people’s beliefs about vision. Research into visual metacognition is exploring how people not only fail to notice or attend to most of a scene, but always overestimate their ability to remember what they have seen. That is, we massively overestimate our powers of perception, attention and memory, and we similarly overestimate other people’s powers of visual cognition.

In his paper entitled “Is the Visual World a Grand Illusion?” Alva Noë summarises the significance of the change blindness work as follows:

The fact of change blindness is widely thought to have several important consequences. First, perception is, in an important sense, attention-dependent. You only see that to which you attend. If something occurs outside the scope of attention, even if it's perfectly visible, you won't see it. In one study, perceivers are asked to watch a video tape of a basketball game and they are asked to count the number of times one team takes possession of the ball (Neisser, 1976; Simons & Chabris, 1999). During the film clip, which lasts a few minutes, a person in a gorilla suit strolls onto the centre of the court, turns and faces the audience and does a little jig. The gorilla then slowly walks off the court. The remarkable fact is that perceivers (including this author) do not notice the gorilla. This is an example of inattentional blindness. Second, perception is gist-dependent. Some changes, for example, in the features that affect the gist of the scene, are more likely to be noticed (Simons & Levin, 1997). Third, it seems that the brain does not build up detailed internal models of the scene; that is, it doesn't perform the integration of information across successive fixations, contrary to the assumption of traditional orthodoxy (Blackmore et al., 1995; Rensink et al., 1997; O'Regan et al., 1999; Rensink et al., 2000; Noë et al., 2000). Or if it does, we have little easy access to this detail. If we did, then presumably we'd keep track of change better than we do.[7]

The “basketball and gorilla” experiment to which Noë refers has now become part of the popular folklore of change blindness, and has been repeated on network television to the general dismay of unprimed viewers. Daniel Levin and Daniel Simons embarked on a series of increasingly surprising change blindness experiments in which subjects failed not only to notice changes of clothes and furniture in a scene but even failed to notice if one actor was switched for another. There was, at the time, some scepticism as to whether these change blindness results could be achieved using subjects in real world situations (as opposed to on video). In response to this scepticism, Levin and Simon developed a variety of situations where subjects were exposed to a quick substitution of their conversation partner (see fig.5). Levin comments:

We still found that about half failed to detect this change. In one case, an initial experimenter approached a subject on a college campus and asked for directions to a building on campus. Midconversation, two other experimenters carrying a door walked between the subject and the first experimenter. While the subject's view was briefly blocked, one of the experimenters carrying the door traded places with the