The text below will bring up an animation. Just look at it once — no cheating! A picture will flash for about a quarter of a second, followed by a color pattern for a quarter second. Then the screen will go blank for about one second, and four objects will appear. Use the poll below to indicate which object (#1, 2, 3, or 4) appeared in the picture.
I’ll let you know which answer was correct at the end of the post, but this test approximates the procedure of an experiment conducted by Kristine Liu and Yuhong Jiang, designed to measure the capacity of visual working memory. If we get enough responses, we can see if our results match theirs.
Previous studies have had conflicting results, with some indicating we can remember a large number of details of a briefly presented scene, and others suggesting that we don’t notice differences between scenes even when we look at them for several seconds (see this Cognitive Daily article for one example).
Liu and Jiang noticed that one problem with some studies is that they asked participants to recall the names of objects, or when they asked participants to choose among possible objects, it was easy to guess. Liu and Jiang avoided these problems by including two entirely different sets of objects in each scene they used. In their first experiment, viewers briefly glimpsed one scene with 10 different objects in it, then had to choose the objects that were in the scene from a set of 20 objects. The 10 “incorrect” answers were the correct answers for the other scene. Half of the viewers saw each scene, and the results were combined:
Responses were better than chance for only the first object selected — just 75 percent accuracy. Apparently a brief glimpse of a scene isn’t enough to introduce many details into working memory.
But perhaps viewers simply forgot the other objects while they were working on the choosing task. To address this issue, a new experiment was designed corresponding to the above animation. A scene was flashed for 250 milliseconds, then viewers had to choose between four possible objects. One object was correct, one was a different object of the same type, and the other two were different objects of the same type (though still plausible for the scene — and as before, one of these objects was used in a second version of the scene). A second group of viewers was allowed to view the image as long as they desired before moving on to the memory quiz. In addition, some of the time viewers saw the scenes with a background context (as in the animation above), and sometimes they were just shown isolated objects on a white background. Here are the results:
When images were briefly flashed in context, accuracy was just barely better than the 25 percent chance rate. Accuracy improved noticeably when viewers saw the objects on the white background. Those who had unlimited viewing time (they viewed the scenes an average of 13.7 seconds) were very accurate, and they, too, improved when they saw the objects on their own.
Liu and Jiang used these results to calculate how many objects, on average, were retained in visual working memory. For the display with context, people can remember about 0.67 of 10 objects when briefly displayed, and 5.33 out of 10 when they have unlimited time. When the objects are shown without a background, these numbers jump to 2.1 and 7.41.
Liu and Jiang argue that their results demonstrate that there are at least two ways our visual system perceives the world: a “fast” route which uncovers the gist of a scene and a “slow” route through which we can process specific details.
So were we able to replicate their experiment? See for yourself — the correct answer to the poll is 2.
Liu, K., & Jiang, Y. (2005). Visual working memory for briefly presented scenes. Journal of Vision, 5, 650-658.