Abstract
The visual span for reading is defined as the number of adjacent letters that can be correctly identified without eye movements, and may impose a sensory bottleneck on reading speed. In this study, we extended the concept of visual span to Chinese characters. We investigated the effect of complexity on the size of the visual span for alphabetic and Chinese characters, and explored the sensory factors contributing to the differences we observed. Perimetric complexity was used to quantify the complexity of binary character images and classify them into four complexity groups (26 characters per group), including lower case alphabetic letters and three Chinese character groups. Six normally sighted bilingual subjects participated. Subjects were asked to recognize trigrams (3 adjacent characters) presented randomly at 17 positions left and right of fixation. Visual span profiles were constructed as recognition accuracy vs. character distance from fixation. The size of the visual span was quantified as the width of the profile (number of characters) at 80% correct. As complexity increases visual span decreased from 10.7 characters for alphabetic characters to 5.0 for the most complex Chinese characters. We also measured the visual span for isolated characters. By comparing recognition profiles for isolated characters and trigrams, we estimated the effects of acuity (decline in recognition away from fixation for isolated characters), crowding (difference between the profiles for isolated characters and trigrams) and mislocations (correct recognition at the wrong position within a trigram). We found that crowding is the major factor limiting the visual span and significantly increases with complexity. There is a smaller effect of mislocation, which is greater for Chinese character recognition. We conclude that visual spans for Chinese characters are narrower than the visual span for alphabetic letters, primarily due to increased crowding associated with the greater complexity of the Chinese characters.
Meeting abstract presented at VSS 2013