Reading words of out order: Comparing serial and parallel accounts of the transposed word effect

A fundamental issue in reading research concerns whether words are encoded serially (one at a time in the order they are encountered) or in parallel (concurrently with the encoding of neighboring words). This debate has reignited recently with evidence showing that readers can encode word order flexibly and often fail to detect grammatical violations produced by transposing adjacent words (e.g., "The white was cat big"). These transposed word effects have been accommodated under both parallel accounts (due to the simultaneous processing of multiple words) and under serial accounts (due to temporal overlap in the integration of adjacent words). However, the precise nature of the effect remains contentious. To examine these contrasting accounts, we conducted six novel experiments that manipulated spatial and temporal aspects of transposed-word stimulus processing. Experiments 1a and 1b show equivalent transposed-word effects for displays that allow or disallow simultaneous processing of transposed words. Experiments 2a and 2b show enlarged transposed-word effects when whole sentence stimuli are displayed at fast presentation rates, but only when spatial cues to word location are unavailable. Experiments 3a and 3b show that the transposed-word effect is larger when the first than second transposed word is presented at a faster presentation rate. Overall, our findings are inconsistent with claims that flexible word order encoding is attributable to parallel word recognition. Instead, they show a temporal component to flexible word order encoding, consistent with claims that they can be explained by overlapping integration stages within serial models.