Abstract
A word is unreadable unless its letters are separately identifiable. Taken from the recent article of Pelli, Farell and Moore (2003), this sentence underlines the crucial importance of letter identification for visual word recognition. Congruently, a fundamental purpose in the study of visual letter recognition is the discovery of the features responsible for accurate letter identification. In the last three decades, researchers have proposed sets of individual features, which predict relatively well letter identification performance and letter confusions. However, these descriptions are usually based more on intuitions regarding the information underlying letter similarity or on confusion matrices than on empirical data directly assessing the information used to recognize letters accurately. In the present study, we aim to reveal the potent features (Gosselin & Schyns, 2002) mediating uppercase and lowercase letter identification in Arial font letters. For this purpose, we explicitly determined the portions of each individual letter which drives its accurate identification. Six participants each identified 26,000 uppercase and lowercase Arial font letters (for a total of 312,000 trials) sampled in image location and spatial frequency by Bubbles (Gosselin & Schyns, 2001). Separate analyses for each individual letter revealed the potent features for uppercase and lowercase Arial font letter identification. The results show that high spatial frequencies support the identification of features that discriminate among visually similar letters (e.g., ‘O’ and ‘Q’ in uppercase). In contrast, low spatial frequencies carry information about the features that are shared among subsets of visually similar letters. These observations are discussed in relation to a letter identification model in which low spatial frequencies are processed initially to determine the subset of the alphabet the target belongs to. Then, high spatial frequencies are processed for information allowing unique letter identification.