Abstract
Recent studies of visual short-term memory (VSTM) suggest that people may remember one of two types of visual representations from a set of objects that recently disappeared. One type of representation is feature-based, and contains separate, unbound features for each object (e.g., one contains an object's color, another its location). The other type is object-based, and contains all the features of one object bound together. Debates persist about how VSTM supports these two types of representations. We report a new study in which we investigate how long each representation lasts after a stimulus disappears.
We conducted an experiment using a change-detection task with colored squares as stimuli. We varied the retention interval between the two displays from 1–4 seconds to test for decay of memory for the first display. Additionally, we included two types of changes. On new-color trials, the colors of two squares would change to two new colors, and on switch trials, the colors of two squares would switch places. Changes on new-color trials can be detected if any color information has been retained, but changes on switch trials can be detected only if bound color-location information has been retained. Thus, the amount of feature-based and object-based information can be estimated separately by performance on these two types of trials.
We found that the amount of feature-based information stayed constant at about 1.4 items' worth over 4 seconds, whereas the amount of object-based information shrank from 2.4 to 1.7 item's worth between 1–4 seconds. These results suggest that the feature-based information stored in VSTM is more durable than object-based representations.
This work was supported by a CSUEB Faculty Support Grant to DF.