Investigating Transposed Word Effects in Reading

Abstract

According to the recently established OB1 Reader model (Snell, van Leipsig, et al., 2018), word position coding is noisy, and words are assigned to sentence positions via a spatiotopic sentence-level mechanism on the basis of length cues and syntactic expectations. A critical prediction that follows from this assumption is that readers should sometimes fail to detect an ungrammaticality created by transposing two adjacent words within a sentence. In line with this prediction, Mirault et al. (2018) initially found that readers are worse at judging the (un)grammaticality of sentences containing a word transposition only, versus sentences containing a word transposition and a syntactically illegal final word – a Transposed-Word effect. Since then, word transpositions have become increasingly relevant to the investigation of whether lexical processing is serial or parallel. The present thesis was aimed at investigating how word transpositions may influence moment-to-moment online cognitive processing during silent sentence reading via eye-tracking. The first experiment explored the influence of word transpositions and final word ungrammaticality on eye movements and grammaticality decisions. The second experiment explored the effects of word transpositions in the parafovea versus the fovea on eye movements and grammaticality decisions via the boundary paradigm (Rayner, 1975). The third experiment investigated the influence of word transpositions in comparison to letter masks in the parafovea on eye movements via the boundary paradigm as well. Across the three experiments, readers exhibited no sensitivity to the presence of a word transposition in parafoveal vision. However, from Experiment 1 and Experiment 2, it was clearly evident that upon fixating the first of the two transposed target words in a sentence, readers experienced significant and immediate disruption to processing. Furthermore, findings from Experiment 2 and Experiment 3 indicated that readers were sensitive to a preview change created by changing the order of the two target words in parafoveal versus foveal vision. Importantly, that sensitivity was likely due to the visual and orthographic rather than the syntactic mismatch between preview and targets and was driven primarily by the invalid preview of the first rather than the second target word. Overall, the results from the three experiments suggest that words are lexically processed serially and sequentially rather than in parallel. Moreover, readers do indeed sometimes fail to detect word transpositions, however, that is likely not driven by the processing mechanisms proposed by the OB1 Reader model. Consequently, current serial attention shift computational models of reading need to be expanded and modified in order to account for why readers fail to detect word transpositions under serial lexical processing assumptions.