What Is Change Blindness? Missing Large Changes in Plain Sight

The standard laboratory version shows a photograph, interrupts it with a brief blank, and shows a modified version, alternating the two. Observers typically take a long time to find changes that are obvious once seen, including large objects appearing, disappearing or changing colour. The interruption is essential, since without it the change produces a motion signal that attention detects automatically, and any disruption serves, including a blank screen, a flicker, an eye movement, a cut in a film or a physical obstruction passing between the observer and the scene. Real-world versions have been more striking still, with experimenters swapping the person a participant was speaking to during a brief interruption and a substantial proportion of participants failing to notice, which has been replicated in several forms.

The results constrain accounts of what visual experience contains:

• •The impression of a detailed stable scene is not supported by a detailed internal representation of it
• •Detail appears to be available on demand by looking rather than held in memory continuously
• •Attention is required to register a change, and unattended regions are not monitored
• •The visual system does not store enough between glances to support comparison across an interruption
• •People are confident they would notice such changes, which is measured and is wrong, and that confidence is itself a finding
• •Related failures occur for unexpected objects in plain view when attention is occupied elsewhere

A separate and equally striking effect concerns unexpected objects rather than changes. Participants asked to count events in a video frequently fail to notice something conspicuous and unexpected passing through the scene, with the best-known demonstration involving a person in a costume walking through a ball game, and roughly half of observers missing it entirely while attending to the counting task. The effect is called inattentional blindness and it differs from change blindness, since nothing changes and the object is simply not noticed. Both indicate that attention rather than exposure determines what is consciously seen. The two effects together have been used extensively in discussions of eyewitness reliability, of driving and of any situation where a person is expected to notice something while occupied with something else.

The findings are frequently overstated and the limits are worth stating. They do not show that vision is unreliable generally, since people navigate, catch objects and recognise faces perfectly well, and the failures occur in a specific situation involving an interruption and an unattended region. They do not show that nothing is retained between glances, since gist, layout and the identity of objects are retained even when details are not. They do not show that people are inattentive in general, since the same mechanisms that produce the failure are what allow attention to be allocated usefully at all. The honest statement is that visual experience feels more complete than the underlying representation is, and that attention determines which parts of a scene are actually processed in detail, which is a specific claim rather than a general scepticism about perception.

The practical implications are substantial and the applications are specific. Driving is the clearest case, since a driver occupied with a conversation or a device may look directly at a hazard without registering it, which is the basis of the finding that hands-free phone use is not meaningfully safer than handheld, since the problem is attention rather than hands. Medical imaging shows the same pattern, with radiologists missing conspicuous abnormalities when searching for something else, demonstrated in studies inserting obviously anomalous objects into scans. Aviation and process control have designed around it. Eyewitness accounts are affected. Film editing exploits it, since continuity errors pass unnoticed across cuts, which is why they survive into finished films. And security screening and inspection tasks suffer directly, which has driven work on how to design such tasks to reduce the effect.

Substantial changes go unnoticed when they coincide with any brief interruption, including an eye movement or a film cut, and real-world versions have swapped the person a participant was talking to. The findings indicate that a detailed internal representation is not maintained and detail is fetched by looking. People are confidently wrong about whether they would notice, which is itself the measured result.