Word order effects in sentence reading

Abstract

The SEAM model (Rabe et al., 2024) and the OB1-Reader model (Snell, van Leipsig, et al., 2018) suggest that readers lexically process words in parallel, with the OB1 model further specifying that those words are formed into a sentence-level representation irrespective of their order of presentation. The serial model, E-Z Reader (Reichle, 2011), in contrast, stipulates that words are identified serially and sequentially. The current eye tracking experiment investigated whether, how frequently, and how rapidly readers detect sentential anomalies arising from word transpositions and ungrammatical sentence final words. We also assessed the consequences in the eye movement record of processing such transpositions and ungrammaticalities to evaluate theoretical claims extrapolated from different eye movement models. This was done via target word pair (transposed vs. non-transposed) and a final word grammaticality (grammatical vs. ungrammatical) experimental manipulations. Readers were better at judging the grammaticality of sentences containing both a word transposition and an ungrammatical final word than those with solely a word transposition. Critically, transposed words caused significant disruption to reading, but not prior to readers fixating the first word of the transposed word pair. Furthermore, an ungrammatical sentence-final word attracted readers’ fixations and caused increased re-reading in the absence of a word transposition compared to when it was preceded by a transposed word pair. Together the results show the importance of canonical word order for natural undisrupted reading and question claims for parallel lexical identification in relation to eye movement control during reading.