Bisection performance indicates spatial word representation

From reading errors of a neurological patient, Caramazza and Hillis concluded: (1) that word length is part of the cognitive representation of words; (2) that the spatial reference frame of this representation is centered on the word; and (3) that this representation is orientation-invariant [Nature (London), 346 (1990) 267-269]. To test these three hypotheses about normal cognition, healthy subjects bisected either lines with word flankers or English words. The length of word flankers systematically affected line bisection, supporting the first hypothesis. In conflict with the remaining hypotheses, however, word bisection was biased toward their beginnings for horizontal words but not for vertical words. A similar bias with pronounceable non-words (but not with meaningless letter strings) and an increase of the word bisection bias in non-native readers suggest that overestimation of the spatial extent of word beginnings is related to lexical access. An attentional scaling hypothesis is proposed that also accounts for a similar eye fixation bias in normal reading.