Representation of matrix patterns in long- and short-term visual memory

Abstract

Two experiments examined two-alternative recognition for matrix patterns either immediately after a brief display (short-term visual memory) or following visual interence (long-term visual memory). The matrix patterns used were constructed so that in some condition could be based on the identification of familiar letter shapes embedded in the patterns (semantic classification). In other conditions recognition involved discrimination between different shapes belonging to the same category, or two unfamiliar shapes, a task requiring precise spatial description of the target. When interference conditions were blocked, long-term recognition was highly dependent on semantic classification of the patterns. Short-term performance was based on shape descriptions of the patterns, rather than on semantic classification, although performance was generally higher when the target pattern contained a familiar shape. When interference conditions were randomized, both long-term and short-term recognition showed similar dependence on the semantic classification of patterns. These results suggest that: (a) the components of visual memory use different methods for coding matrix patterns, and (b) the type of information maintained by the short-term component is variable and can be selected under voluntary control.