On the spatial limits of parallel word processing in reading

Various models of reading assume that information from up to five words is processed in parallel. Although there is evidence that foveal words can be processed simultaneously with directly adjacent words, it remains to be seen whether three words is the limit. To empirically test this, we designed a lexical decision flanker task with three flankers on each side of the target. In two experiments (offline (N = 49) and online (N = 98)), target words were either orthographically unrelated to all flankers or repeated in one out of six flanker positions. Stimuli were briefly presented, allowing us to assume that flanker effects, if any, would stem from simultaneous rather than sequential processing of the target and flankers. We observed flanker effects for flankers immediately adjacent to the target word. However, the relatedness of flankers in more remote positions did not impact recognition of the target. Our results suggest that word processing occurs for approximately three words in parallel, which is more than what some theories (e.g., E-Z Reader, Chinese Reading Model) suggest, but less than what some other theories (e.g., Glenmore, OB1-reader) suggest.