Abstract
There is great interest in designing fonts and electronic methods of display to increase reading speed, but so far, the improvements have been modest, a few percent. Visual span is the number of letters that can be identified with a single fixation. It is approximately 10 letters, centered on fixation. Letters beyond this are too crowded to recognize. As a result, saccades must be executed to the next set of letters. If crowding can be reduced so that more letters can be identified with each glimpse, reading speed should increase proportionally. Crowding is weak when target and flankers are dissimilar. Consequently, text that alternates in color from letter to letter (black-white-black) should increase reading speed. Yet this does not work (Chung & Mansfield, 2009). In a previous experiment we found that the alternation leads to grouping of letters, which increases crowding instead of reducing it. In an attempt to reduce crowding we ensure that letter-letter dissimilarities do not form a pattern. Observers read a page of text as quickly as they can while maintaining comprehension. White text is presented on a gray screen. Using a gaze-contingent, a length of text (slightly larger than estimated visual span) centered on current fixation is rendered black. The outermost black letters should now be uncrowded by the surrounding white letters. This should increase the span size and result in increased reading speed. Indeed, we find that subjects read about 18% faster using our gaze-contingent paradigm [t(3)=13.8837, p<0.001] than with unaltered text. We manipulate various stimulus factors (within the gaze-contingent window) such as color, size, and contrast, all with similar effects. We conclude that with a gaze contingent display, reading speed can be increased.
Meeting abstract presented at VSS 2012