@article{c3702f8aa8db4c9f96255552ad3994d5,
title = "Better early than late for a filler: An fMRI study on the filler-gap order in language",
abstract = "The neural substrates of syntactic movements have been heavily investigated; however, little attention was paid to the fact that there was a cross-linguistic preference for filler-before-gap (filler-first) to gap-before-filler (gap-first) structures in subject-verb-object (SVO) languages. This fMRI study aimed to explore whether there was a cognitive basis for such a preference. Different filler-gap orders resulting from various syntactic movements were tested in Chinese: topicalization (filler-first) and relativization (gap-first, including subject and object relative clauses). The data showed that gap-first vs. filler-first contrasts activated the left anterior temporal gyrus (L-ATL) and sometimes the left thalamus. We argued that the L-ATL (and the left thalamus) was recruited because deeper semantic retrieval was performed on the verb to facilitate its merge with the gap for thematic role assignment. Our results provided a possible cognitive explanation for the preference for filler-first to gap-first structures in SVO languages.",
keywords = "ATL, Filler-gap order, Relative clause, Syntactic movement, Topicalization",
author = "Lee, {Ting wu} and Chan, {Shiao hui}",
note = "Funding Information: The recruitment of the left temporal lobe, especially the L-ATL, confirmed our main hypothesis that processing gap-first structures incurred a greater cognitive cost than processing filler-first ones and that such a cost might have to do with the deeper semantic retrieval of the verb. Also, this brain activity was not modulated by the different gap-filler distance between SRC and ORC, further supporting our hypothesis that this retrieval should be related to the verb and its prediction about the adjacent gap (and thus the potential filler) and that the distance between the gap and the filler was irrelevant.The further retrieval of verb-specific information in gap-first structures is also supported by our unexpected observation of the involvement of the left posterior thalamus (pulvinar) in ORC-TOP. The left thalamus has been found to play a special role in semantic processing (Hebb & Ojemann, 2013). In particular, the pulvinar is connected with cortical areas, such as the temporal cortex, prefrontal cortex, and inferior parietal lobule (Gutierrez et al., 2000), and the left pulvinar is invoked in semantic searching tasks, particularly mediating the synchronized firing of cortical regions that encode parts of an object concept (Assaf et al., 2006). Therefore, it is likely that the left pulvinar was recruited to search further when a verb had to be merged with a semantically empty gap. Note that the absence/presence of the pulvinar activation for SRC-TOP/ORC-TOP may be attributed to the relationship between the gap and the verb. From a theoretical point of view, the object is known as the “internal argument” that directly merges with the verb to form a larger constituent (verb phrase), while the subject is the “external argument,” with its role being assigned by the verb phrase (Grimshaw, 1990; Williams, 1981a, 1984). Therefore, the elements of the merging between a subject gap and a verb in SRC might be slightly different from those between an object gap and a verb in ORC. Future study is needed to confirm this speculation.A further note about the comparison between SRC and ORC is necessary. While it is generally agreed that head-initial SRCs (i.e. the head noun preceding the RC) are easier to process than ORCs in English and other languages (Frauenfelder et al., 1980; Frazier, 1987; Gouvea, 2003; King & Just, 1991; Mak et al., 2002; Schriefers et al., 1995; Traxler et al., 2002), the SRC advantage in head-final RCs (the head noun preceded by the RC) as in Mandarin Chinese is rather controversial (SRC advantage: Lin, 2006; Gibson & Wu, 2013; Vasishth et al., 2013; Ueno & Garnsey, 2007; Xiong et al., 2019; ORC advantage: Hsiao & Gibson, 2003; Carreiras et al., 2010; Packard et al., 2011; Wang et al., 2017). Our result showed that the processing of SRC and ORC in Mandarin Chinese did not induce differential brain activity. On the one hand, our finding is in line with a previous ERP study, which found that both the sustained anterior negativity and P600 were detected for long and short wh-dependencies, suggesting no effect of filler-gap distance on working memory load (Phillips et al., 2005). On the other hand, our finding is inconsistent with previous findings supporting either SRC or ORC advantage. We believe such inconsistency could be attributed to how the data were analyzed. In this study, we only analyzed the BOLD signals associated with the relative clause and the head noun (i.e. Frame 1 in Table 1), not with the whole sentence as in previous research (Bulut, 2018; Constable et al., 2004; Rogalsky et al., 2008; Xu et al., 2020). It is likely that the SRC or ORC advantage associated with BOLD signals may not appear until much later (i.e. Frame 2 or 3 in Table 1), but since our main focus was on the filler-gap order and we did not control the content of Frame 2 in SRC and ORC to be identical, the exploration of the advantage issue between SRC and ORC is beyond the scope of this study.This work was supported by the National Science and Technology Council (formerly known as the Ministry of Science and Technology), Taiwan, R.O.C. under the grant number MOST 108-2410-H-003-020-MY3. We also thank the Imaging Center for Integrated Body, Mind, and Culture Research for their assistance in data collection. Funding Information: This work was supported by the National Science and Technology Council (formerly known as the Ministry of Science and Technology ), Taiwan, R.O.C., under the grant number MOST 108-2410-H-003-020-MY3 . We also thank the Imaging Center for Integrated Body, Mind, and Culture Research for their assistance in data collection. Publisher Copyright: {\textcopyright} 2023 Elsevier Ltd",
year = "2023",
month = may,
doi = "10.1016/j.jneuroling.2023.101126",
language = "English",
volume = "66",
journal = "Journal of Neurolinguistics",
issn = "0911-6044",
publisher = "Elsevier BV",
}