Letter-By-Letter Readers: Evidence for Parallel Processing

Recognition of written words by normal subjects is massively parallel, at least when they are presented undistorted with the letters in their normal order in free fields. Recent models of the processes of word recognition (e.g., McClelland & Rumelhart, 1981) have therefore assumed that all the visual elements making up a word are processed simultaneously and in parallel (see Fig. 1). This assumption is based on a series of results showing that the effects of word length on oral reading time or accuracy are small: a range of studies reviewed by Henderson (1982) show effects of between 6 and 63 ms per letter, but as Schiepers (1980) demonstrates, such effects can be eliminated entirely when words are presented at the fovea.