4.2. fNIRS
Overall, the fNIRS analysis revealed a limited prefrontal activity in SRC and more widespread prefrontal activity in ORC condition. Moreover, direct ORC>SRC comparison revealed significant activity both in the left and right DLPFC. Given that the ORC processing was expected to be more difficult than SRC processing in Turkish, this finding was consistent with the initial hypotheses.
In the SRC condition, the middle frontal gyrus (i.e., BA 9) showed significant hemodynamic activity (Fig. 3A). It was found that left BA 9 was active during maintaining or remembering the piece of information just received (Raye et al., 2002). Therefore, activity in these regions can be due to the subjects’ need to recall the relationship between the object and subject of the main and relative clauses to answer the follow-up questions correctly. Furthermore, Jörgens et al. (2007) showed that BA 9 is also active during a sentence completion task. The other regions showing significant activity were in the lateral portions of the left prefrontal cortex including BA 46. It was shown that left BA 46 shows activity during semantic processing (Ni et al., 2000; Wang et al., 2008) and syntax processing (Ni et al., 2000) in sentence level. Even though the focus of the present study was to investigate syntax processing at the sentence level, subjects had to process the semantics as well to answer the questions. Furthermore, it was shown that the left DLPFC plays role in various aspects of sentence processing, such as parsing or plausibility (Hertrich et al., 2021).
As can be seen in Figure 3B, the strongest hemodynamic activity was mostly localized in the left DLPFC (i.e., BA 9, 45, and 46) in ORC condition. It was shown that the left DLPFC involves in language processing (Hertrich et al., 2021; Klaus & Schutter, 2018) and syntax processing (Chen et al., 2006; Indefrey et al., 2001; Just et al., 1996; Ni et al., 2000). Therefore, it can be said that the activity in the lateral portions of the left prefrontal cortex occurs due to sentence processing. The other active region was the right DLPFC, covering mostly BA 46 (Fig. 3B), which was also reported for relative clause > noise processing (Hassanpour et al., 2015) and ORC>SRC processing (Kovelman, Baker, et al., 2008). It was suggested that the right DLPFC could be an indicator of greater cognitive control required for processing more difficult sentences (Ding et al., 2021). It is known that both left and right DLPFC are part of the executive control network (Hertrich et al., 2021) and the right DLPFC, together with the anterior cingulate cortex, plays role in error/conflict monitoring and resolution (Badre & Wagner, 2004; Jackson et al., 2022). It was also shown that the right DLPFC is active while processing more complex garden-path sentences (den Ouden et al., 2016).
Several explanations have been put forward for RC processing asymmetry in Turkish (Özge et al., 2010). The authors’ first claim was that SRCs are more common in Turkish compared to ORCs, which can lead to SRC processing advantage. The other account was based on the word-order similarity. In Turkish, the canonical word order is as SOV, while ORCs are produced with SVO, and SRCs are produced with OVS word order. As can be seen, the word order of SRC shows higher similarity to the canonical word order of SOV, which can lead to SRC processing advantage in Turkish. The perspective shift was proposed as another explanation for SRC processing advantage. Relative clauses are prenominal structures in Turkish and the focus in SRC is on the agent (subject), while the focus in ORC is on the patient (object). Since the canonical sentence starts with a subject in Turkish, it was expected that starting of a sentence should have the agent role. If it does not coincide with this role, as in ORC containing sentence, then a conflict arises (i.e., shift in perspective is required) and leads to a processing disadvantage. As explained above, SRC and ORC are constructed with different suffixes. Özge et al. (2009) stated that the suffix used in ORCs (i.e., “-dık”) could also be seen in other structures in Turkish, such as noun complement clause (Kahraman et al., 2010), while the suffix used in SRCs not. Therefore, it could be said that a direct form-function mapping leads to a processing advantage in SRCs in Turkish, while participants should employ further cognitive resources such as error monitoring to correctly process ORCs.
Direct ORC>SRC contrast revealed a significantly greater hemodynamic activity in three frontal regions (Fig. 4). The first region corresponds to the left DLPFC (i.e., BA 45 and 46). The left BA 45 corresponds to Broca’s Area, known to be involved in language processing. Furthermore, multiple imaging studies on English relative clause processing have reported that the activity in BA 45 and BA 46 increases in more difficult condition (Caplan et al., 2002; Chen et al., 2006; Just et al., 1996; Kovelman, Baker, et al., 2008). An EEG study carried out in Basque, in which SRC was expected to be the more difficult condition, revealed that SRCs induced greater P600 amplitudes compared to ORCs, especially in the left dorsolateral regions (Carreiras et al., 2010). Another study showed that the left DLPFC activity could be associated with checking the thematic roles in the sentence as well as with response selection among alternatives (Kovelman, Baker, et al., 2008). The left DLPFC was also associated with working memory (Hertrich et al., 2021), which was also crucial process in the present task as the subjects had to identify the agent and patient in both main and relative clauses to answer the question.
The other active channel collects signal from both left frontopolar cortex and orbitofrontal cortex (i.e., BA 10 and 11, respectively). BA 10 and 11 have not been reported in RC literature, except for one study (Caplan et al., 2008). The authors reported that BA 11 shows significant activity for ORC>SRC comparison. However, the locus of this activity was more in the lateral part of the prefrontal cortex. On the other hand, the active channel in this report was located in the medial parts of the prefrontal cortex. Therefore, it is more likely that the source of this activity was the frontopolar cortex. Even though this region is not specifically associated with language processing, it was shown that it involves many cognitive processes including problem-solving and decision-making (Koechlin & Hyafil, 2007). They proposed that the frontopolar cortex involves in cognitive branching processes, such as computing a complicated math problem by first computing smaller fragments and then combining the results of these fragments. This might also be the case in relative clause processing, during which the participants should identify the role of each noun phrase in both main and relative clauses and then compare them to answer the question correctly. Another study also stated that thematic role control is mostly performed by frontal areas of the left hemisphere (Kovelman, Baker, et al., 2008). Even though we made changes in the stimuli to prevent subjects to take advantage of thematic roles, it is hard to rule out the possibility that subjects employed this strategy in syntactically more difficult sentences.
Lastly, we found a significant activity difference for ORC>SRC in the right DLPFC (i.e., BA 45 and BA 46). The activity of this region was associated with processing more complex sentences (Ding et al., 2021; Kovelman, Baker, et al., 2008). As stated above, the right DLPFC is part of the domain-general executive control network and plays role in many cognitive processes including error/conflict monitoring. It could be the case that the activity in right DLPFC was associated with the perspective-based explanation. To recall, Turkish speakers expect an agent at the beginning of a canonical sentence. On the other hand, encountering an ORC at the beginning of the sentence creates a conflict to be resolved, resulting in greater activity in this region compared to the SRC condition. The other possibility could be that the activity within this region may reflect the evaluation of the alternatives induced by the suffix “-dık” used in ORCs. It was claimed that the right DLPFC shows activity in response selection (Badre & Wagner, 2004). As stated above, the suffix “-dık” is used in other structures in Turkish and the participant should eliminate the alternatives to choose the correct representation, which might require further cognitive control and conflict monitoring/resolution.