Abstract
Reading is considered an inefficient activity that requires independent detection of individual letters without integrating word or sentence information. We studied the use of word- and sentence-shape information using word discrimination and reading tasks with spatially distorted text in 18 naïve observers. Threshold durations were measured as a function of the wavelength and magnitude of spatial distortion applied to text for word/non-word classification of 4 and 8 letter words, and true/false classification of 4-word sentences. Spatial distortion slowed word recognition and reading. Unlike the uni-modal tuning functions observed in noise masking studies, we find multi-modal tuning functions that correspond to maximal impairment at around 2 cycles per letter, 2 cycles per word and 2 cycles per line of text. These data imply that information at these word- and sentence-length dependent scales are important for word recognition and reading, This finding is consistent with the robustness of reading letter-substituted text, proof-reading errors and the ineffectiveness of letter-based assistive reading methods.
Meeting abstract presented at VSS 2015