Developing Intelligence

Change Blindness and Attention

If a large object were to suddenly disappear from your field of view, you might expect that you would notice its disappearance. However, change detection research has demonstrated that we have a surprisingly poor ability to detect even large changes to a visual scene (see here and here for examples).

Skeptics might complain that this “change blindness” could simply result from absent-mindedness: maybe you happened not to notice the changing feature. A 2002 article by Becker and Pashler actually rules out this explanation – indicating that our internal visual representations may indeed be as meager, fragile, and temporary as the change blindness phenomenon suggests.

To confirm that lack of attention cannot explain change blindness, the authors had nearly 50 subjects view a circular array of 12 digits. After subjects announced either the highest digit present in that display (in the first experiment) or the lowest digit not present in that display (in the second experiment), the screen went blank for 150 msec. Then a second array was presented, which was either the same as or different from the previous display; if different, only a single digit had changed.

Subjects were correct over 95% of the time in the highest & lowest digit tasks (demonstrating that they had indeed viewed the digits in the display) but were actually quite bad at detecting when one of the previously viewed digits changed (with only around 33% correct). This performance is equivalent to that obtained when subjects only need to detect a change, rather than having to perform both tasks. Therefore, change blindness cannot be explained merely as a result of failing to attend to the feature or item that changes!

On the other hand, subjects were quite good at detecting when the previously-announced number changed (as you would expect.) The authors suggest this reflects the fact that change detection is dependent on visual short term memory rather than attention alone. Consistent with this account, 33% accuracy is exactly what would be expected if subjects can only maintain 4 digits from the original display for later comparison; 4 objects is widely considered the average capacity limit for visual short-term memory.

Becker and Pashler conclude that internal visual representations do not consist of a “cumulative record” of previous experiences, and thus are not limited by the number of objects to which we have previuosly attended. Instead, these representations are limited by short-term memory capacity limits. This is consistent with the claim that our internal representations are incredibly sparse, as compared to the subjective experience of a rich and detailed internal visual representation. Therefore, inattention alone cannot explain change blindness.

Related Posts:
Neural Processing and Familiarity
A New Mode of Perception and the followup: Mindsight Reconsidered
Our Ability to See Change Over Time


  1. #1 BenP
    January 30, 2007

    What have change in the picture with the plane ? I don’t get it !

  2. #2 Chris Chatham
    January 30, 2007

    Hi Ben – it’s close to the center of the picture…

  3. #3 Foxy
    January 30, 2007

    Good lord … I stared at that plane for an hour without seeing that.

    It would be interesting, I think, to figure out what kinds of things and objects people instinctively focus on, when trying to solve a problem like that.

    I assume of course that people, knowing the trick, would pick certain elements to focus on, but who’s to say.

  4. #4 Paul
    January 31, 2007

    Something which might be interesting in relation to your excellent post: possibly the most renowned video example of this effect. To describe the video would be to ruin the effect (provided it has not been seen before): the aim for observers is to count how many times the basket-ball has been passed between players of two teams (two teams, two basket-balls – count the passes):

    Link to video:

  5. #5 lindsey
    February 1, 2007

    I, too, had a very hard time seeing the diference in the plane images. What I find intriguing is the fact that once I know what the difference is, I can’t *not* see it when I view it again, even if I look in the periphery. (This must be why juries can’t understand how a radiologist could have missed a lesion on an X-ray, once it’s pointed out to them.) I also remember an example shown at a science museum of an apparently random set of black and white blobs. After applying an overlay, it’s obviously a black and white cow, even after the overlay is removed, and the image cannot be seen as random again.