Chapter 6 Explaining two-picture change blindness

For this unit, the term change blindness refers to the failure to notice changes in animations that alternate between two pictures of a scene. Later we will also talk about other situations in which people miss changes, but this chapter focuses on the two-picture alternation animations. In previous years, you probably already saw some of those amazing demonstrations. Here, however, we will learn somewhat different lessons than what you learned before.

First you need to realize that when we view a scene, we typically remember very few details about it. That’s true even when we actively try to memorize the contents of the scene. When watching this movie, please scrutinise the scene carefully.

Wasn’t that amazing? In some ways, we humans are a lot dumber than we think! People usually don’t notice any of the several changes made to the scene. We’ll circle back to this Whodunnit movie, but first let’s talk about a case that may be even more striking.

6.1 Blindness for gradual changes

In the Whodunnit movie, the changes occurred off-screen, when the camera was focused tightly on the detective on the left. One might expect that if the changes happened right in front of your eyes, you would notice them.

Amazingly, we fail to notice changes right in front of us, too, if they happen very gradually, as illustrated in this movie.

6.1.1 The “grand illusion of visual experience”

Most people are surprised by the blindness for changes in the gradual-change movie. Many researchers were very surprised, too, and some came to the conclusion that there is a grand illusion of visual experience. This is the claim that while people think that they are simultaneously experiencing the whole visual field, they are wrong about that - it is an illusion. These researchers explain change blindness with the claim that at any one time, you are only experiencing a small portion of the visual field, parts that you are particularly attending to. In other words, these researchers claim that visual experience is subject to a strong bottleneck.

However, this conclusion that there is a bottleneck on visual experience may be premature. To understand why, we need to consider in more detail what the failure to notice changes might mean. We need to consider the processing that’s needed to detect a change.

6.1.2 What is needed to detect a change?

Let’s consider what it takes to detect the change of an object or part of a scene:

  1. An internal representation of that object that is different before and after the change.
  2. A process that compares what was identified earlier to what is being identified now.
  3. A process that calls attention to, or brings into conscious awareness, the instances of change.

Apparently, at least one of the above three processes is lacking. Let’s consider #1 first. It is the case that all incoming retinal signals across the scene get processed. Unfortunately, however, if the object is in the periphery, the retinal signals may not be high-resolution enough for the representation to be different before and after the change. This is because vision is low resolution in the periphery (4).

For many real-world scenes, then, #1 above is sufficient to explain why people don’t notice changes. However, #1 is not enough to explain all failures to notice changes. Even when researchers create displays in which all the objects are big enough and widely-spaced enough to see in the periphery, still people miss many changes. For example, the changes are large enough in some classic demonstrations like this boat scene.

So, #2 or #3 or both are lacking. This is likely due to a bottleneck. These processes are limited in capacity, so they cannot simultaneously process all objects in the visual scene. And what about the “grand illusion of visual experience?” Well, it seems quite possible that we may have experience of objects without having processes that correspond to #2 and #3. In other words, the conclusion that there is a grand illusion of visual experience may be a hasty one (Noë, Pessoa, and Thompson (2000)). When people are surprised by change blindness, their mistake may be failing to realise that there’s various processes required to notice a change, and visual experience may not always involve those.

Only a finite number of neurons can fit in our head, and evolution seems to not have prioritized processing of #2 and #3. The brain has not devoted neurons to constantly comparing what you’re seeing now to what you saw half a second ago. Comparing what was present at two different times requires the limited resources of attention to be at that location at the two different times. We don’t know why evolution did not prioritize these, but one possibility is that a full comparison process (#2) would require a lot of neurons, and animals like us have been able to get by with other, simpler processes, which we will discuss next.

6.2 Bottom-up attention and flicker/motion detectors

While limited capacity means we can’t fully process the whole visual scene simultaneously for changes, brains have evolved some simple tricks that help us catch many changes. One of these is that our brains have flicker or motion detectors that do simultaneously process every part of the scene.

At my home, mounted high in the corner of the carport, is an inexpensive motion detector that you can buy at the hardware store. This device is wired so that if it detects motion, the carport light comes on. There is nothing fancy about the processing within it - not much circuitry is required.

When done by neurons, too, crude motion detection doesn’t require much work or energy (you can learn more about this in PSYC3013). In one or more of the visual retinotopic maps located in our brains, each bit of the map has flicker/motion detectors sitting there that ordinarily fire as soon as something happens in the scene. Specifically, sudden disappearance or sudden appearance of an object will make these flicker/motion detectors fire.

Firing of those flicker/motion detectors can call attention to a location. This is an instance of bottom-up attention (5). Thus, the brain uses bottom-up attention as a work-around: the flicker or motion ordinarily caused by a change summons attention to a location, and then more limited-capacity processes work out what’s changing there.

In summary, we have evolved to process simultaneously across the scene only a few things. Two of these things are flicker and motion. Thus, detecting motion and flicker is NOT capacity-limited. We rely on this to signal the locations where something is happening.

Very gradual changes do not trigger our flicker/motion detectors. But when changes are sudden, this will stimulate our motion or flicker detectors, which in many circumstances will call attention to the associated location.

These facts about the brains of humans and other animals are one reason that animals stay very still when they are worried about predators. Thanks to their camouflage, many animals can be hard to notice when they’re not moving, but as soon as they move, they’re quite conspicuous (watch this) and predators’ attention goes straight to them.

However, what happens if motion or flicker occurs in multiple places? It won’t be clear which of the associated locations attention should go to. O’Regan, Rensink, and Clark (1999) have demonstrated how this can enable changing blindness with a display feature that they called “splashes” - see Traffic with splashes and Traffic without splashes. The idea is that these movies might resemble the situation if splashes of a puddle hit your windshield while you are driving - the splash would trigger your motion detectors and thereby call your attention, preventing your attention from going the location of potentially-important other changes.

Broader background motion can also present a problem - if everything in the scene is moving, then our motion detectors are stimulated everywhere and attention may not go to the location of a change.

Now you can fully understand why people take a long time to find the change in the classic change blindness animations.

6.3 Two-picture change blindness

Most of you have seen animations like that of the boat scene or this Paris scene, which sandwich a blank screen in between the two versions of the picture. It’s like a blank screen sandwich! (The two pictures of the scene are analogous to the two chocolate biscuits and the ice cream is analogous to the blank screen).

Here is a schematic of the timeline of a blank screen sandwich, wherein the arrow represents time.

Two pictures of a scene are alternated, with a blank screen between them

That blank screen is critical - it creates flicker everywhere in between the two frames. That is, when the picture of the scene is replaced by the blank scene, it creates a flicker signal everywhere, and then flicker everywhere again when the second scene comes on.

When the blank screen is removed from a blank screen sandwich, the scene change is conspicuous; this animation is an example. In it, the only location that tickles your transient detectors is that of the change. As a result, your attention goes straight to the location of the change.

Without the blank screen, the only location of flicker was the location of the changing object. The flicker called your attention to that location. With the blank screen, there’s flicker everywhere, so there is no indication of which of the many locations contains the change.

6.4 Searching without a clue

When the blank screen is in the animation, the flicker/motion detectors provide no clue as to the location of the change. You might think that in this situation, people would search about randomly, or perhaps in a systematic fashion, something like searching from left to right and then top to bottom; in 5 you learned that at least in some circumstances, the attention of many people moves about approximately in reading order.

Researchers have collected data on this for scenes such as the dinner date change blindness scene. Specifically, the eye movements of people were recorded while they were viewing these animations, to see where people look. These eye movements provide a pretty good indication of where people direct their attention. In a previous chapter (4), you learned that most movements of attention are overt - if people are interested in something, they usually will look right at it.

Pink lines indicate the trajectory of eye movements made by people searching for a change.

The above image shows some eyetracking data. The long straight lines represent big jumps of the eyes from one place to another as the participants tried to determine what was changing. As you can see, the eyes dwell mostly on the couple’s faces, their hands, and some objects on the table. So, the locations that people looked were not random at all, and frequently do not occur in left-to-right reading order.

You already knew that unique colors and other features are salient and attract attention, but here you can see that other properties of a scene also affect attention. Two thousand years ago, Aristotle wrote that “Man is by nature a social animal.” People are very interested in people, and in working out what they’re thinking and feeling. Things like people, bodies, and objects like food and wine are sometimes referred to as a scene’s high-level properties. The word “high-level” is used in part to indicate that those properties take more processing to extract, and thus are represented at later (“higher level”) stages of the brain compared to, say, color.

To understand the meaning of this scene involves working out the facial expressions of the people and how they are interacting with each other, based on their postures and the objects in front of them. Many participants were so captivated by this social stuff that these participants never looked at some of the other objects, like the railing behind the couple (which was what was actually changing in the change blindness animation). Have you ever gone “people watching?” If so, you might be like one of those participants (this is just a thought, NOT a validated psychological finding).

The above image shows the data for only one particular scene. Does the pattern of fixating the eyes more on face and bodies hold for other stimuli as well? Rigby, Stoesz, and Jakobson (2016) investigated this issue. Sixteen participants watched twelve four-second movie clips and twelve still-frame images from several episodes of a TV show that was heavy on dialogue and characters (the Andy Griffith show). The soundtrack was turned off during the viewing.

Average amount of time spent looking at different parts of the scene.

The results, shown above, provide further support for the hypothesis that attention is biased towards faces. In another study, Rösler, End, and Gamer (2017) flashed pictures of scenes for just a fifth of a second, so people had time for only one eye movement, and found that people disproportionately looked at parts of the scene with faces or bodies.

These social biases of attention are used by web app and advertisement designers who seek to control what you attend to. You may have noticed that many ads have a picture or animation of a person in them, even when this is completely unnecessary and superfluous to the information provided. It’s kind of hack of your attentional system to get you to read or watch ads.

There is some evidence that the attention of many children with autism spectrum disorder (ASD) is less biased towards faces than is that of typically-developing children. The study whose results are plotted above was one study that investigated this, by also including among their participants a group of sixteen adults with autism spectrum disorder.

Average amount of time spent looking at different parts of the scene, in sixteen adults with (right) and without (left) autism spectrum disorder.

Based on this study, children with ASD don’t spend as much time looking at faces. What would you expect, then, for the pattern of performance in change blindness in people with autism spectrum disorder? Kikuchi et al. (2009) conducted a change blindness experiment and varied whether what changed was the head of a person, another object, or a change to the color of the background.

A schematic of one of the trials in the experiment. This trial is an example of the head change condition. The head of the person to the left was replaced by another head.

The blank screen sandwich was looped until the participants pressed a key. The participants were then required to report what the change was, by pointing at it or with a verbal description. As one measure of performance, the researchers examined only those trials where the participants correctly detected the change and plotted the response time on those trials. Shorter response times indicate that the person detected the change faster.

Mean correct response time for detecting a change. The black line represents children with ASD and the dashed gray line typically developing children. Error bars are one standard error.

Figure 6.1: Mean correct response time for detecting a change. The black line represents children with ASD and the dashed gray line typically developing children. Error bars are one standard error.

6.5 Exercises

Answer these questions and relate them to the first four, and the last, learning outcome (2):

  • Why can classic change blindness animations be described as a “blank screen sandwich?”
  • Why are gradual changes hard to detect?
  • What effect do splashes and other irrelevant sudden changes in a scene have on our ability to detect important changes? How do they have that effect?


Kikuchi, Yukiko, Atsushi Senju, Yoshikuni Tojo, Hiroo Osanai, and Toshikazu Hasegawa. 2009. “Faces Do Not Capture Special Attention in Children with Autism Spectrum Disorder: A Change Blindness Study.” Child Development 80 (5): 1421–33.
Noë, Alva, Luiz Pessoa, and Evan Thompson. 2000. “Beyond the Grand Illusion: What Change Blindness Really Teaches Us about Vision.” Visual Cognition 7 (1-3): 93–106.
O’Regan, J. Kevin, Ronald A. Rensink, and James J. Clark. 1999. “Change-Blindness as a Result of ‘Mudsplashes’.” Nature 398 (6722): 34–34.
Rigby, Sarah N., Brenda M. Stoesz, and Lorna S. Jakobson. 2016. “Gaze Patterns During Scene Processing in Typical Adults and Adults with Autism Spectrum Disorders.” Research in Autism Spectrum Disorders 25 (May): 24–36.
Rösler, Lara, Albert End, and Matthias Gamer. 2017. “Orienting Towards Social Features in Naturalistic Scenes Is Reflexive.” PLOS ONE 12 (7): e0182037.