Every utterance has a possibility, to a greater or lesser degree, of threatening the addressee’s face, that is, their public self-image maintained in society (Brown and Levinson, 1987), and in turn can jeopardize the interpersonal relationship between the addresser and the addressee. In particular, the act of refusing an invitation is a face-threatening act (FTA) in that the refuser causes a serious conflict with the inviter’s friendly attitude. In order to redress the conflict, the refuser should use an appropriate politeness strategy to refuse indirectly, namely, by leaving the sentence unfinished to obscure their true intention of not performing the invited action. Although native speakers (NS) can easily understand polite strategies of indirect refusals, non-native speakers (NNS) often have trouble understanding such strategies as they are not fully accustomed to conventionally polite ways of hiding one’s true intention (Kasper, 1984).

The realization of indirect speech acts differs depending on the syntactic properties of each language. The salient characteristics of the Japanese language are: it has a head-final structure in which arguments of the subject (S) and the object (O) necessarily precede the head (i.e., verb: V); and in a compound sentence, a conjunction comes at the end of the preceding clause as shown in (1a), contrary to head-initial languages (e.g., English, Chinese) in which the conjunction comes at the beginning of a clause.

As in (1a), the preceding subordinate clause ending with the conjunction typically indicates the reason for refusal, while the subsequent main clause includes the true intention of the refusal itself. This syntactic property allows the Japanese language to easily omit the latter main clause including the negation of the invited action, such as ikenaidesu (meaning “I cannot go”) in (1b), resulting in an unfinished sentence with only a subordinate clause that ends with the conjunctive particle -kara (meaning “since”).

Although it is quite common in many languages to convey an FTA of refusal indirectly by only stating the reason not to do the invited action, as in (1b), it is characteristic of Japanese to leave the sentence unfinished by putting the conjunctive particle -kara (“since”) at the end of the sentence, although the particle syntactically requires the subsequent main clause. Japanese NS are known for their strong preference for this strategy of omitting the main clause, including their true intention, in order to indirectly realize FTAs in everyday conversations (Maynard, 1989). Notably, it is assumed that such unfinished sentences are not ambiguous for NS by virtue of conventionalization (Taguchi, 2014). Uttering such unfinished sentences can then be interpreted as a universal politeness strategy of performing an indirect speech act in order to be conventionally indirect (Brown and Levinson, 1987). However, NNS who are not fully accustomed to the conventionalized politeness strategy may find syntactically unfinished sentences that lack the main clause stating the speaker’s true intention of the FTA to be anomalous.

Experimental pragmaticists have demonstrated the complex linguistic and cognitive processes involved in the real-time comprehension of how conventionalized indirect speech acts are realized. Studies using electroencephalograms (EEG) indicate that NS recognize the intention of a speech act at a very early stage (Egorova et al., 2013; Gisladottir et al., 2015). Evidence from functional magnetic resonance imaging (fMRI) studies suggests that making inferences about indirect speech acts requires greater cognitive load, since such processing induces higher activation of regions of the brain, including the bilateral prefrontal cortex (Shibata et al., 2011; van Ackeren et al., 2012), the temporoparietal junction, and the medial prefrontal cortex (e.g., Bašnáková et al., 2013; Schurz et al., 2013; Feng et al., 2017), which are related to mentalization, cognitive empathy, or theory of mind (ToM), all of which are interchangeably used to refer to the capacity to understand other people by attributing mental states to them (Baron-Cohen, 1988). Although neurocognitive evidence for NNS’ processing of indirect speech acts is scarce, several behavioral studies suggest that it is difficult for NNS to comprehend indirect speech acts. Reaction time data showed that English NNS were not as fast or accurate in processing indirect requests and refusals as were NS (Taguchi, 2005). Holtgraves (2007) also indicated that NNS’ speech act recognition was not as reflexive as that of NS. However, these previous studies exclusively focused on European head-initial languages with SVO word orders. The influence of the syntactic structure of non-European head-final languages with SOV orders on the comprehension of indirect speech acts in NS and NNS remains unknown.

Utilizing high temporal resolution, event-related potentials (ERPs) of EEG provides useful findings for understanding the time course of online cognitive processing for the syntactic and pragmatic realizations mentioned above. From a syntactic viewpoint, the well-known ERP component of P600, which positively peaks approximately 600 ms after the given event, has been observed for the syntactic processing load across many languages, not just by NS but also by NNS (e.g., Frenck-Mestre et al., 2009; Gillon Dowens et al., 2010; Lemhöfer et al., 2014; Meulman et al., 2014). Regarding Japanese language processing, increasing numbers of ERP studies demonstrate higher cognitive load for complex syntactic processing in terms of word order (e.g., Yano et al., 2019) and syntactic island constraints (i.e., movement constraints on syntactic objects; Tokimoto et al., 2019). Furthermore, the functional role of the P600 has recently been required rethinking, such that it reflects integration difficulty in linguistic contexts rather than the interpretation of syntactic structure (e.g., Brouwer et al., 2012; Xiang et al., 2016), but these studies did not include Japanese. In order to differentiate the neural basis of pragmatic processing such as politeness judgments and that of syntactic processing, we need to consider the interplay of multiple ERP components, which depends on the cognitive demand of pragmatic and syntactic processing.

A variety of pragmatic realizations is known to induce another popular ERP component of negativity, N400, peaking at around 400 ms after stimulus onset. Although the N400 has traditionally been assumed to be an index of various kinds of semantic/pragmatic integration, such as inferring non-literal meanings (Pynte et al., 1996), retrieving conceptual information from context (see Canal and Bambini (2021) for a review), and irony comprehension (Katz et al., 2004), a recent parsimonious hypothesis of single-stream account for language processing suggests that the N400 reflects the retrieval of lexical information from memory, while the P600 relates to semantic/pragmatic processing instead (Brouwer et al., 2012). We should also notice that a robust N400 effect can be observed not only with visual stimuli, but also with auditory stimuli. This is relevant to the present study, which focuses on neural substrates for the pragmatic processing of indirect meanings of FTAs conveyed in spoken conversations. For example, a recent study reported an auditory N400 effect for processing non-literal meanings (Kessler et al., 2021). Another recent investigation compared the pragmatic N400 effect between English NS and the proficient NNS (Jankowiak et al., 2021), showing similarities between the two groups when comprehending figurative meanings. Taken together, these results indicate that the N400 effect should serve as a valid index reflecting both NS and NNS’ cognitive load when inferring indirect meanings from unfinished sentences of refusal within the auditory modality.

We may assume that if one speaks an unfinished sentence to indirectly realize an FTA in order to be polite to the addressee, such an unfinished sentence should mitigate the negative emotion postulated to be aroused in the addressee by virtue of FTA which could cause conflicts between the speaker and the addressee (e.g., Chen, 2013). In this context, NS are expected to promptly perceive such a politeness strategy, which might be reflected in an earlier neural response. Indeed, several neurolinguistic experiments of NS show that emotional processing via verbal information elicits earlier ERP components, such as early posterior negativity (EPN). Several studies have reported the EPN effect on emotional language processing (Schacht and Sommer, 2009; Mittermeier et al., 2011; Kissler and Herbert, 2013). With regards to Japanese, Kiyama et al. (2018) demonstrated the emotional effect of EPN on mood-modulating sentence-final particles. They found an individual difference in the EPN in terms of mentalization or ToM, indicating that Japanese NS with lower mentalizing ability yielded a greater EPN for complex processing of linguistic sentence-final particles that implicate the speaker’s emotions. The enhanced EPN in people with lower mentalizing ability may reflect their hypersensitivity to complex emotional linguistic markers which require flexible interpretation of the speaker’s attitudes. This suggests that low mentalizers are less able to grasp subtle indications of emotions, due to their stereotypical knowledge of these markers. Nevertheless, there is still a lack of neurolinguistic comparisons between NS and NNS concerning individual differences in emotional reactions to language.

Individual differences in mentalization during the processing of unfinished sentences of indirect refusals can also be expected because such sentences require the addressee’s ability to infer the speaker’s motivation and emotion for omitting the main clause indicating the intention of refusal. Several studies have indicated that participants with lower mentalizing ability took greater cognitive load for comprehending indirect requests (Trott and Bergen, 2019) and linguistic expressions in non-literal context (McKenna et al., 2015). This finding is in accordance with the increasing number of clinical pragmatic studies investigating syntactic and pragmatic (i.e., non-literal language processing) atypicality in adults and children with autism spectrum disorder (ASD), a diagnosis based on persistent deficits in mentalization or ToM and by restricted, repetitive patterns of behavior and interests (for a review see Cummings, 2014). Linguistic atypicality in people with ASD in Asian cultures is particularly evident in the processing of bound morphemes like sentence-final particles that have no substantial meaning but indicate the speaker’s subtle moods or emotions (e.g., Kamio, 1994; Matthews and Yip, 1994; Koo and Rhee, 2013).

In sum, this study investigates the neural substrates for perceiving sentences left unfinished to realize a conventionalized indirect speech act of refusal in Japanese, a head-final language. It focuses in particular on the effects of individual differences in mentalization. Specifically, we examine the following four hypotheses: The first hypothesis (H1), which focuses on syntactic unfinishedness, assumes that the P600 effect would be elicited by unfinished sentences that lack the main clause indicating the intention of refusal only in NNS. NS would not show a syntactic ERP effect because they automatically appreciate this highly conventionalized form of syntactic incompleteness, as it is fully conventionalized. The second hypothesis (H2) concerns the pragmatic processing of indirect refusal. We predicted that the N400 effect would be greater in NNS than in NS for unfinished sentences of indirect refusals as an FTA relative to finished sentences of direct refusals, since NS sub-unconsciously process such conventionalized indirect refusals. As for the third hypothesis (H3), which concerns the emotional reaction to the language form, we assumed that, compared to NNS from different cultural backgrounds, NS familiar with Japanese conventions would elicit earlier ERP components for emotional processing, such as EPN for finished sentences of direct refusals than for unfinished ones of indirect refusals. Lastly, with regards to our fourth hypothesis (H4), we expected that people with lower mentalizing ability would evoke greater amplitudes of the ERP components mentioned in the first three hypotheses, given lower mentalizers’ hypersensitivity to emotional language processing, as suggested in a previous study (Kiyama et al., 2018). The first three hypotheses entail a group comparison between NS and NNS, while the fourth requires a post hoc analysis to examine the effect of individual differences on the findings of the group comparisons. Conversely, these hypotheses concerning differences between NS and NNS would not be supported if the NNS have, not just advanced linguistic abilities such as precise syntax and a large vocabulary, but also a great aptitude to quickly learn the implicit pragmatic conventions of the Japanese language.

The present ERP study examined four hypotheses to compare NS and NNS as well as the individual difference in mentalization for processing of indirect refusals conventionalized in Japanese. As a result, the neural substrates for perceiving unfinished sentences to perform a conventionalized indirect refusal as an FTA have been elucidated, even though the subjective assessment for the unfinished sentences is not significantly different between the two groups. H1 concerning syntactic unfinishedness, H2 for pragmatic processing of indirectness of the refusal, and H3 regarding emotional reaction to the unfinishedness were generally supported, while H4 was only partially supported.

First, as expected in H1, we observed the P600 effect for processing unfinished sentences compared to finished sentences only in NNS, whereas this effect was not found in NS. Although the P600 effect has been consistently reported for syntactic anomalies in Japanese (Tokimoto et al., 2019; Yano et al., 2019), the present study found the P600 to be prevalent for NNS’s processing of unfinished Japanese sentences in indirectly realizing a speech act of refusal. This suggests that NNS of Japanese particularly experience a processing difficulty with this kind of unfinished sentence. Conversely, it is obvious that the NS, who did not yield the P600, may process these sentences in a different manner. The finding that the NS may not consider the unfinished sentences anomalous is consistent with the previous theoretical assumption that Japanese NS are accustomed to syntactic unfinishedness in face-threatening situations, as they prefer to obscure the true (impolite) intentions. The difference in the P600 effect between NS and NNS supports NS of Japanese’s automatic processing of unfinished sentences ending with a conjunctive particle (-kara meaning “since”) in the situation of refusals.

Particularly, our NNS participants were native speakers of Chinese, which has a different syntactic property from Japanese. In Japanese, the structure of a subordinate clause ending with a conjunction is commonly used, while the structure of the Chinese subordinate clause requires a conjunction coming at the beginning of a clause (Yap et al., 2014). This difference reflects whether the language is consistently head-initial (like Chinese) or head-final (like Japanese) at all phrasal levels. It is known that head-final languages typically have case-marking information, which enables incremental structure building in a language with flexible word orders (Kazanina, 2017). Given a large number of psycholinguistic experiments (e.g., Kamide, 2008; Sato et al., 2013; Kurumada and Jaeger, 2015; Witzel and Witzel, 2016) demonstrating that NS of Japanese as a head-final language have an advanced syntactic ability of anticipatory sentence processing, Japanese NS may be predicting the contents of the latter main clause of a complex sentence much earlier, while they are presented with the preceding subordinate clause. Chinese, contrarily, is categorized as a head-initial language without case markers and flexible word order, which may influence the processing difficulty of Chinese learners of Japanese inferring the omitted main clause of complex sentences in Japanese.

Consistent with H2 with regard to pragmatic processing, the N400 effect was elicited for the indirect speech act of refusal compared to the direct speech act only in NNS. This finding supports the notion that NNS with different cultural conventions need more effort to understand indirect speech acts. Even if they are highly proficient in the target language, NNS should rely on explicit verbal cues to recognize the speaker’s true intentions. The present experimental design prepared the pragmatic control condition of FSd as a direct speech act of refusal by adding the main clause stating the direct refusal meaning (e.g., “It’s impossible”) after the preceding subordinate clause stating the reason to refuse (e.g., “I’m busy that day”), where the NNS participants could use the main clause to specify the speech act. However, the target condition of the USi lacked the main clause indicating the refusal intention. Presumably, the lack of the main clause causes NNS to experience difficulty inferring the illocutionary force of a refusal. NS, contrarily, did not show the enhanced load of processing of the indirect speech act of refusal, suggesting that they have no extra cognitive load for indirect speech act recognition. The finding may reflect that these unfinished Japanese sentences of indirect refusals are highly conventionalized in NS of Japanese.

Furthermore, within the same time window of 300–500 ms, another unexpected positive component was induced for the same comparison of the USi with the FSd as a pragmatic control. Although caution is advised, this positivity may be related to the P300, which is known to be evoked by the oddball paradigm (e.g., Polich and Margala, 1997; Katayama and Polich, 1999; Herrmann and Knight, 2001), which detects attention for improvable stimuli compared with provable ones. Especially in the linguistic vein, the oddball P300 is reported to reflect the mental representation for context updates (Donchin and Coles, 1986), in that newly given improvable linguistic contexts require updating the acquired mental representation for provable contexts. In light of this, the USi that evoked P300 compared to the FSd may result from an unexpected flow of dialog for NNS, presumably because they experience a kind of context mismatch in which they prefer a direct sentence with an explicit refusal, rather than an indirect sentence obscuring the true intention.

Concerning H3 for emotional processing for whether the indirect refusal is finished or unfinished, only NS elicited significant ERP effects of both positive and negative deflections for the FSi compared to the USi. However, NNS did not yield any earlier ERP effects. We interpret the NS’s significant negativity as EPN in accordance with H3, and the positivity as P200, another well-known early ERP index for emotional processing, because the P200 is reportedly dependent on the emotional congruency of words within the context (León et al., 2010; Leuthold et al., 2015), and because the effect is known to respond to auditory language stimuli as well as visual stimuli (Paulmann et al., 2008). These NS’s ERP effects for the FSi in earlier time window suggest their negative emotional reaction to letting sentences finished, given that NS of Japanese are especially reluctant to complete sentences themselves. For NS, such finished sentences may act as less polite and less conventionalized strategies for performing an indirect refusal. Rather, they clearly value unfinished sentences, as the politeness theory (Brown and Levinson, 1987) argued that uttering unfinished sentences functions as a kind of politeness strategy. NNS who are not fully accustomed to Japanese culture, on the other hand, may not be able to distinguish the politeness functions between finished and unfinished sentences. This NNS’s insensitivity to Japanese politeness might have resulted in no significant effects concerning emotional ERPs, but it yielded a syntactic anomaly (i.e., P600) for the forms at the sentence-final position.

Lastly, in terms of H4 for individual differences in mentalization, we found a correlation between AQ and the amplitude of N400 only in NNS’s processing for perceiving indirect refusals. Namely, NNS participants with higher AQ (reflecting lower mentalizing ability) evoked larger amplitudes of the N400 for unfinished sentences to convey the intention of indirect refusal, compared to finished sentences as direct refusals, whereas NS did not indicate such correlation. The result suggests that NNS with lower mentalizing abilities require more cognitive loads to infer the indirect intention of unfinished sentences of face-threatening refusals. NS, on the other hand, did not indicate individual differences in mentalization for any ERP effects for syntactic, pragmatic, and emotional processing. This was inconsistent with the previous ERP study (Kiyama et al., 2018), which found that NS’s individual mentalizing ability modulated the effects of the emotional component of EPN for processing Japanese sentence-final particles depending on context. The present study’s manipulation of sentence-final expressions may not be distinctive enough for NS of Japanese, because the preceding context was identical across the three conditions of USi, FSi, and FSd, unlike Kiyama et al. (2018).

Overall, the present study found opposing ERP effects for unfinished sentences of refusal between NS and NNS. Namely, NS seem to smoothly process the unfinished sentences and feel anomalous with finished ones, while NNS may experience a higher syntactic and semantic cognitive load for unfinished sentences. The unfinished sentences utilized in this experiment were syntactically incomplete, ending with a conjunction -kara (meaning “since”) without the following main clause. The finding that NS did not experience a syntactic anomaly despite this unfinishedness supports the supposition that some conjunctions in contemporary Japanese do not necessarily function as conjunctions, but are often perceived as a kind of mood-expressing sentence-final particle, as the conjunctions are placed at the end of the sentence (National Institute for Japanese Language and Linguistics, 1951; Haugh, 2008). This ongoing syntactic/semantic status change of conjunctions in Japanese are not limited to the particle -kara, but also other particles including -node (meaning “since”), -shi (meaning “and”), -kedo (meaning “but”), presumably reflecting Japanese NS’ reluctance to make strong assertions through the single speaker finishing a sentence (National Institute for Japanese Language and Linguistics, 1951; Haugh, 2008). It has long been assumed that Japanese NS prefer to leave sentences unfinished, expecting the addressee to complete them, so that the speaker and the addressee collaborate to make a sentence together. This phenomenon is known as co-construction (Mizutani, 1980). Indeed, many qualitative linguistic studies of conversation analysis observe the complex turn-taking patterns in Japanese conversations, including co-constructions and long, twisted sentences allowing the addressee’s overlapping inputs (Tanaka, 1999; Hayashi, 2003). Even though the style may be inherent in Japanese culture, it is unsurprising that NNS have particular difficulty following such chaotic structures of conversation. Here lies potential for misunderstanding between Japanese NS and NNS, such that NS may feel uncomfortable with NNS’s finished sentences and might unfairly perceive them as impolite. This study provided the first neurophysiological evidence concerning these potential misunderstandings resulting from Japanese preferences for unfinished sentences, serving as a reference especially for adult learners of Japanese who have already acquired their own pragmatic language skills inherent in their native cultures.

Nonetheless, the current study contains several limitations, which require further examination in future studies. First, as our initial goal was to differentiate the neural responses to the syntactic unfinishedness in conveying refusals as an FTA between NS and NNS, we focused on the effect of the syntactic properties of head-directionality of languages, namely, whether each group’s native language was head-initial or head-final. However, the preference for syntactic unfinishedness when realizing FTAs is heavily influenced by different cultural backgrounds, besides the syntactic properties per se. As indicated by the dichotomy between individualism and collectivism (Hall, 1976; Triandis and Gelfand, 1998), it is assumed that people from individualistic cultures emphasize self-consciousness, while those from collectivistic cultures focus on interdependence in interpersonal interactions. These differential patterns of behavior and thinking would determine whether those peoples prefer to leave sentences unfinished in order to perform indirect speech acts (Searle, 1969) to impress polite addressees. Although our participants were Japanese NS and NNS whose native language was Chinese, both within the alleged collectivistic cultures, future research should compare these effects with NNS from individualistic cultures. Of course, the realization of refusal in actual encounters depends on how the preceding invitation was conveyed. To better understand the neural substrates for processing refusals, we should investigate the neural responses to the speech acts such as invitation and request preceding the refusal, which are significantly influenced by the cultural characterization as well (e.g., Holtgraves and Joong-Nam, 1990 for request; Kiyama et al., 2019 for invitation). The next concern is about the NNS’s effect on individual differences regarding the attitude of mentalization among typically developed young adults utilizing AQ as a convenient screening questionnaire. To confirm the current finding about the correlation between individual AQ and the N400 effect reflecting the pragmatic anomaly for the unfinished indirect sentences of refusals, we need to expand studies of the neural basis for indirect speech acts to include people with ASD. Naturally, we should differentiate the interaction effect of autistic traits or mentalizing ability with the sex difference, in the light of an apparent male bias in ASD prevalence (e.g., Werling and Geschwind, 2013). Further, the present ERP experiment was limited to a relatively small number of participants (i.e., n = 28 for each group) and item number of stimulus dialogs (i.e., 30 sets). Particularly, the less sufficient number of participants could obscure the tendency of individual differences. Hence, we should exercise caution in our current interpretation of the ERP findings, particularly concerning its correlation with the individual difference in mentalization. Further replication studies will contribute to better understand more plausible ways of influences of individual differences in cognitive preferences on the comprehension of indirect speech acts.

This electrophysiological study first demonstrated that the neural response to unfinished sentences to perform a conventionalized indirect speech act of refusal differs between NS and NNS of Japanese, which is representative of head-final languages. NNS, who are not accustomed to the syntactic property, appear to find the syntactic structure anomalous and experience an increased cognitive load when processing unfinished sentences, implying an indirect refusal in a conventionalized manner. Further, the NNS’s individual mentalizing ability seems to boost their sensitivity to the pragmatic anomaly of indirect refusal in an unfinished sentence, presumably because higher mentalizers have sensitive awareness of sentence-final expressions in addition to the proposition of a sentence. Conversely, NS of Japanese inherent in the cultural convention did not induce such pragmatic cognitive loads for unfinished sentences when performing a conventionalized indirect refusal. Overall, these findings provide evidence that a syntactic anomaly inherent in a cultural convention and individual mentalizing ability play an important role in comprehending an indirect speech act of face-threatening refusal.

The datasets generated during the current study are available from the corresponding author on reasonable request.

The studies involving human participants were reviewed and approved by the Ethics Committee of the Graduate School of Arts and Letters, Tohoku University, Japan. The participants provided their written informed consent to participate in this study.

MW, ST, GS, MK, and SK contributed to the conception and design of the study. MW, GS, and SK collected the data. MW, ST, TU, and SK ran the data analysis. MW, ST, MK, and SK interpreted the data and discussed the results. MW and SK wrote the manuscript. ST, GS, TU, and MK review the manuscript. MK and SK acquired funding. SK supervised the project. All authors participated in interpreting the findings and approved the content of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.