Change Blindness Blindness



 This page contains demonstrations discussed in the following paper:
Scholl, B. J., Simons, D. J., & Levin, D. T. (2004). 'Change blindness' blindness: An implicit measure of a metacognitive error. In D. T. Levin (Ed.), Thinking and seeing: Visual metacognition in adults and children (pp. 145-164). Cambridge, MA: MIT Press.
The demonstrations are provided as MP4 movies, which can be downloaded or viewed directly in most web-browsers. These movies are a bit large and choppy, but they should be sufficient to illustrate the basic conditions. If the movies seem too choppy, try downloading the movies and playing them off your local hard drive.  
 
Note that these 'flicker' movies must be set to 'loop' in order to function correctly; this is something you can do from your Quicktime player.  
 
Most people have strong but mistaken intuitions about how perception and cognition work. Such intuitions can give rise to especially pernicious 'metacognitive errors', which are directly fueled by visual experience. In this chapter we explored one such metacognitive error, which infects our intuitions about visual awareness and the perception of change. In 'change blindness' studies, observers fail to notice large changes made to displays when they are viewing, but typically not attending, the changed regions. This phenomenon has been the focus of much recent research, largely because it is so surprising: people vastly overestimate their change detection ability. Here we demonstrate and quantify an implicit effect of this metacognitive error, and explore some of the factors which mediate it. Observers viewed an original and a changed photograph which repeatedly alternated, separated by a brief blank interval. They were told that the change could be added to the 'flickering' at any time. In reality the change was added either immediately (Experiment 1) or after 4 s (Experiment 2). Upon detecting the change, observers were informed of their response time and were then asked to estimate when the change had been added. Observers underestimated the degree to which they were 'change blind', typically inferring that the change had been added much later than it actually was. Average estimates ranged up to 31 s after the 'flickering' began -- over 85 times the correct value. Such effects were further magnified in an additional study (Experiment 3) which employed natural scenes and changes specifically designed to induce a high degree of this 'change blindness blindness' (CBB). These studies collectively demonstrate that CBB can persist across many trials in an actual change-detection task and provide a new way to quantify and explore the factors which mediate CBB. This research highlights the extent to which we can overestimate the fidelity of some aspects of visual processing.  
 
This research employed several different types of flickering 'change blindness' displays. Indeed, a main message of this chapter was that the results obtained with such methods will depend greatly on the nature of the images and changes used -- things which are nearly always implemented haphazardly but never reported in detail. We can distinguish between a number of different kinds of trials (many of which owe their existence to Jen Catena's efforts):  
 
Absurdly Easy Change Detection Trial #1 (81 KB)
Absurdly Easy Change Detection Trial #2 (72 KB)  
Some change detection trials may be absurdly easy to detect, because the changes involve the disappearance of salient foreground objects. These trials will yield neither change blindness (CB) nor 'change blindness blindness' (CBB).  
 
Absurdly Hard Change Detection Trial #1 (80 KB)
Absurdly Hard Change Detection Trial #1 (with change cued) (81 KB)
Absurdly Hard Change Detection Trial #2 (74 KB)
Absurdly Hard Change Detection Trial #2 (with change cued) (75 KB)  
Of course, you can make changes much harder to detect, but eventually -- as in these movies -- they lose their appeal. The changes here are just too miniscule: they are nearly impossible to find, but in the end nobody will be surprised by that. (The 'cued' versions of these movies show where the tiny changes are.)  
 
High-CBB Change Detection Trial #1 (132 KB)
High-CBB Change Detection Trial #2 (113 KB)
High-CBB Change Detection Trial #3 (110 KB)
High-CBB Change Detection Trial #4 (143 KB)  
Here, in contrast, are 4 examples of change detection trials which engender high levels of change blindness (just like the previous category), but also high levels of change blindness blindness (very much unlike the previous category). The changes in these movies are big and tend to take up a significant amount of screen 'real estate'. Nevertheless, many subjects will find them tricky to detect (unlike the first category above). Their nature and size, however, gives rise to a big 'Wow!' factor when they are eventually detected. As a result, these trials typically yield high degrees of CBB: once subjects see the changes, they can't believe that they missed them for so long -- and so in these experiments they would estimate that they had been 'added' recently.