Many languages allow flexible word orders, and how humans parse and comprehend them has been one of the central questions in psycholinguistics. Mounting evidence has suggested that not only syntactic factors (e.g., the formation of a filler–gap dependency) but also discourse-related factors (e.g., the discourse status of phrases) influence the processing of non-canonical word orders. However, far too little attention has been paid to subject-initial non-canonical word orders, leaving it unclear how these two factors interact during the processing of such structures. Given this background, we conducted two self-paced reading experiments in Japanese, focusing on inchoative constructions involving a theme subject and a causer adjunct. Experiment 1 showed that the pre-verbal phrase was read more quickly in causer–theme sentences than in theme–causer sentences. We attributed this processing asymmetry to a filler–gap dependency between the sentence-initial theme subject and its original position in the theme–causer sentences. Experiment 2 revealed that this reading-time difference diminished but did not entirely disappear under the condition that a phrase marked as discourse-given appeared first. This pattern aligns well with previous findings concerning the processing of object-initial non-canonical word orders. Taken together, we conclude that syntactic and discourse factors interact in the processing of subject-initial non-canonical sentences, just as they do in the processing of object-initial ones. This study also proposes what we call a sentence-pair method for the examination of the phrase-by-phrase processing of mono-clausal, structurally simple sentences. Thus, this paper makes not only empirical and theoretical contributions but also a methodological contribution to psycholinguistics.
Asami et al. (2025) studied this question.