**DOI**: https://doi.org/10.1080/23273798.2025.2496910 **Abstract** Recent research using a speeded grammatical decision has revealed inconsistent results regarding whether a transposed-word effect can be obtained in response times (RTs) when words are presented serially. Although some studies have reported significantly longer RTs in making ungrammatical decisions to a transposed-word sequence relative to a control sequence, other studies have reported null effects in RTs while obtaining a significant effect in error rates. In a pre-registered replication experiment using the RSVP (rapid serial visual presentation) procedure, we showed that a transposed-word sequence yielded significantly more errors than control sequences, but we failed to find a reliable transposed-word effect in RTs. Our finding was further supported by a subsequent meta-analysis on RT measures which suggested that the overall transposed-word effect under serial presentation is significant, albeit small, and smaller than the transposed-word effect observed under parallel presentation.