The distribution of information within letters

To study the relative importance of various letter segments for letter recognition, we presented each letter of two alphabets, English and Hebrew, preceded by a brief presentation of an intact or a mutilated version of it. Mutilations were done by eliminating a specific segment. It was reasoned that the more critical the eliminated segment is, the less the mutilated version activates the letter code in memory and, thus, the longer it takes to name the subsequently presented target letter. This procedure was successful in detecting significant differences consistent with our expectations. In further analysis, it was shown that the latency data were highly correlated with the distinctiveness of the mutilated segment, its uniqueness over the alphabet, its impact on the letter global shape, and its topography within the letter and other variables. The dependency of latency on the various factors varied considerably between alphabets. Some correlational analyses were done to evaluate the roles of the various factors.