Letter and configuration information in word identification

Abstract

Subjects learned meanings for 16 invented words (e.g., BARDREL) and practiced categorizing each word on the basis of its meaning. During learning and practice, each word appeared consistently in either script or uppercase type. Subsequently subjects categorized both versions of each word (BARDREL and Bardrel) on the basis of meaning. On the first exposure to each word during the test phase, categorization times were about 50 msec slower for unfamiliar versions. The difference diminished rapidly, disappearing after two to six presentations of the previously unfamiliar versions, depending on prior practice. Training transferred better from uppercase to script than from script to uppercase. The results suggest that subjects relied on configuration information, especially for script words, as well as letter identity information.