Exploring metapragmatics of politeness lexemes using a computational approach

1 Introduction

Following the constructivist turn in linguistic studies, politeness has been divided into first-order politeness, which is conceptualized as laypersons’ discursive interpretations of what politeness is and what can be considered polite, and second-order politeness, which is defined as a scientific understanding of this concept (Eelen 2001; Watts et al. 1992). Based on either or both the emic (layperson) or etic (scholarly) perspectives, previous studies have surveyed institutional explanations for politeness lexemes (Haugh 2004) and speakers’ self-reports (e.g., Fukushima and Sifianou 2017; Obana and Tomoda 1994), contributing invaluable insight into the metapragmatics of politeness lexemes across different cultures. However, a potential pitfall remains when both scholars and laypersons cannot provide effective explanations for a politeness lexeme. For example, Fukushima and Haugh (2014) found that their participants cannot always distinguish the three politeness lexemes that they investigate. Culpeper and Haugh (2021) also noticed the infinite regress issue of definition, that is, defining one politeness lexeme requires the definition of another politeness lexeme. Thus, they propose to identify the metapragmatics of a politeness lexeme by analyzing its co-texts that characterize the construal of the lexeme in contexts (Agha 2004; Firth 1957). They use collocation analysis, which is, however, confined to locations or frequencies of the co-texts in the corpus. The current study employs a more sophisticated computational approach to address both the theoretical and methodological challenges. It delves into the metapragmatics of two politeness lexemes that both scholars and laypersons could not articulate clearly.

The politeness lexemes that this study chooses are kongson (‘veneration, politeness’) and kyemson (‘modesty, humility’) in Korean. While politeness in East Asia has drawn continuous attention in its debates with politeness theories developed in Euro-American contexts (e.g., the debate between volitional and conventional politeness), the diversity of politeness expressions within East Asian languages has scarcely been examined (see Fukushima and Haugh 2014; Haugh 2004 for the few exceptions). The Korean language has especially received a rather disproportional amount of attention in comparison to its rich collection of politeness lexemes (Kim 2011). What seems to even hinder the research on Korean politeness lexemes is the difficulties in elucidating their differences. As Chen and Lee (2021) have noted, even L1 Korean speakers struggled to explain kyemson and kongson as well as their differences. Nevertheless, they used these lexemes differently and not interchangeably in context. Scholars also use the two lexemes to cross-define each other (Cho 1980; Hong 2020), falling into the infinite regress of definitions that Culpeper and Haugh (2021) called for to avoid. This study chooses the lexemes and applies a novel methodology to:

1. provide a computational method for the examination of metapragmatics of politeness lexemes for which both etic and emic explanations fall short;

2. provide an insight into Korean politeness lexemes that have not been systematically examined thus far.

In the sections below, we begin by reviewing previous studies on the metapragmatics of politeness lexemes, discussing their methodological approaches and outcomes. We then move on to an introduction of the two lexemes in Korean by situating them in the broad context of East Asia. We explain the corpus data that we extracted and the computational method that we adopted in Section 4. Findings based on the data analysis are presented in Section 5, followed by a discussion and a conclusion that remark on the contributions of this study, and highlight the working direction for future research.

2 Metapragmatics of politeness lexemes

Metapragmatic underpinnings of politeness lexemes have attracted a great amount of researchers’ interest since the divorce between first-order politeness (politeness understood by laypersons) and second-order politeness (politeness as a scientific and analytical concept) (Eelen 2001). We have witnessed that discussions about cultural specificities of politeness shifted from the debate between conventional politeness (emphasized in East Asian contexts; see Gu 1990; Hwang 1990; Ide 1989) and volitional politeness (proposed as universal in Brown and Levinson 1987), and mappings between speech strategies and politeness (e.g., Blum-Kulka 1987; Yu 2011), to the increasing comparisons of metapragmatic understanding of politeness lexemes in different cultures. The investigations into politeness metapragmatics can largely be categorized into three strands by the types of data and analytical methods that they adopted – namely, institutional explanations of politeness lexemes, speakers’ reports, and corpus data.

In the first strand, Haugh (2004) drew on dictionary definitions, scholarly interpretations, and dialogical examples to compare English “politeness” to three Japanese lexemes: teinei, reigi tadashii, and keii hyoogen, that had often been translated into politeness. He found that the three lexemes were associated with different social qualities. Teinei conveyed the demonstration of warmth and kindness to others, reigi tadashii referred to both self-modesty and upward respect for others in accordance with social norms, and keii hyoogen fell between the two, featuring both considerations for others and less upward respect. In contrast, English politeness concerns predominantly the demonstration of a polished self-presentation and considerations shown to others. Similarities between English politeness and the three Japanese lexemes were argued at a higher and more abstract level, summarized by Haugh (2004: 105) as “self-oriented politeness” and “other-oriented politeness”. Similar to Haugh’s (2004) approach, Terkourafi (2011) analyzed didactic literature that taught politeness in Egypt, India, and China. Gu (1990), who proposed principles for Chinese politeness, based his analysis on historical classics and dialogical examples.

Based on analyses of prescriptive texts, the first strand of research contributed to the understanding of politeness norms, but spoke little of politeness in interactions and discursive interpretations of individuals who may or may not adhere to the norms. The shortcoming gave rise to the second research strand of politeness metapragmatics, which focused more on laypersons’ self-reports.

Researchers developed a variety of techniques to collect emic interpretations of politeness lexemes from laypersons, including self-narratives (e.g., Obana and Tomoda 1994), questionnaires (e.g., Fukushima and Sifianou 2017), lexical mapping tasks (e.g., Pizziconi 2007), and metapragmatic interviews (e.g., Fukushima and Haugh 2014). For example, in Pizziconi (2007), lexemes that participants believed to be related to politeness were collected first. Then, the conceptual similarity between politeness and each of these lexemes was evaluated by L1 English and L1 Japanese speakers. The findings pointed out that British politeness was closely related to being “appropriate”, “nice”, “considerate”, and “courteous”, while Japanese politeness was more associated with being “respectful”, “appropriate”, “considerate”, and “sincere”. The seemingly shared value of “consideration” or being considerate also differs across cultures. By comparing questionnaire answers from Japanese female college students and their Greek counterparts, Fukushima and Sifianou (2017) found that the Greek group paid more attention to egalitarian respect as a way of showing consideration, in contrast to the Japanese group, which showed more empathy. Similarly, using metapragmatic interviews, Fukushima and Haugh (2014) studied attentiveness, empathy, and anticipatory inference in Japanese and Taiwan Chinese. They found that attentiveness was understood in relation to respect and courtesy in Japanese, in comparison to propriety or appropriate behaviors in Chinese. Their finding of propriety in Chinese was supported by Su (2019), who used the same technique to compare politeness understood by mainland Mandarin speakers and Taiwanese Chinese speakers.

Based on the metapragmatic explanations provided by laypersons, these studies have contributed invaluable insight into cultural specifics of the seemingly equivalent politeness lexemes in different languages. However, as Fukushima and Haugh (2014: 177) warned, this approach “only gets us so far when considering the possible relevance of such notions to evaluations of im/politeness”. Three main issues hindered the second approach from moving forward, the first being the restriction in sample size and their representativeness. The small number of participants (n = 40 in the case of Fukushima and Haugh 2014), may represent only the selected group of the population, especially considering that the selection was not random. The second issue is methodological and analytical constraints. The measurement tools, such as lexical mapping tasks in Pizziconi (2007), reduced the evaluation of politeness to two-dimensional connections between lexemes. The reduction may oversimplify the actual complexity of any single politeness lexeme. Metapragmatic interviews, albeit being more capable of capturing the complexity, could not address the participants’ own confusion. For example, Fukushima and Haugh (2014) found their participants could not distinguish the three lexemes of their investigation – namely, attentiveness, empathy, and anticipatory inference – just as Korean speakers cannot articulate the difference between kyemson and kongson. The third issue is lack of context, interactional data, and lexemes-in-use. The focus of these studies being on what speakers say about the concept of politeness, poses limitations on understanding what speakers actually do with politeness. As Dunn (2013) criticizes, the investigation of politeness-related lexemes from the metalinguistic approach often relies on decontextualized semantic analysis, which needs to be complemented by considering expressive practices.

A few recent studies, including Culpeper and Haugh (2021) and Oliver (2022), have started to address this issue by investigating politeness-in-use through corpus analysis. Closely related to the methodology of this study, Culpeper and Haugh (2021) analyzed collocations of the term “offensive”, and treated the lexico-grammatical patterns found in its collocations as metalanguage of offence. They compared the metalanguage of British offence to that of Australian English, and found that the British concept of offence has a stronger connection with “abusive”, “insulting”, and “racial” behaviors, in contrast to offence in Australian associated with “inappropriate”, “rude”, and “morally wrong” behaviors.

We should note, however, that collocation analysis only conveys the expressions that occurred right before or after the selected politeness lexeme. In contrast, the co-texts that are relevant to the key politeness lexeme and that are important to describe the context of its usage could occur far beyond the range of collocations. Thus, the corpus analytical techniques have not yet fully explored the potential of analyzing co-texts to explore metapragmatics of politeness lexemes.

Together, the disadvantages of the above-mentioned three analytical approaches have called for more sophisticated methods that can (i) capture a broad range of co-texts, and calculate their importance and relevance to the context where a politeness lexeme is used; (ii) extract reliable patterns from discursive interactions, instead of summarizing prescriptive norms; (iii) include data that are large in size and diverse in contexts to obtain more representative analysis results; (iv) capture the complexity of lexeme construal without reducing the analytical dimensions; and (v) explore politeness lexemes that are actually used in different contexts, while distinguishing them when laypersons and/or scholars have difficulties doing so. In response to the five challenges, the current study adopts a state-of-art topic modeling technique. It allows us to survey more than 10,000 texts where kyemson and kongson are actually used, analyze co-texts in terms of their importance and relevance, present patterned results from discursive usage of the two lexemes while maintaining the diversity of interpretations, and most importantly, distinguish the politeness lexemes that laypersons and scholars could not provide effective interpretations for.

3 Kyemson and kongson in Korean

Kyemson and kongson originate from the Chinese lexemes 谦逊 (qianxun in the Chinese pinyin system) and 恭逊 gongxun. They are Sino-Korean, which is a type of loanword that occupies approximately 60 % of the overall Korean vocabulary (Chang 1996). Kongson has been widely translated into “politeness” (e.g., Brown 2015; Cho 1980; Hong 2020; Tao et al. 2016) and “veneration” (Chen and Brown 2022). Kyemson was translated into “modesty” or “humility” (e.g., Kim and Kim 2004), although Bond et al. (2012) argued that “modesty” was a culturally normative requirement and politeness tactic which does not concern genuinely about humility in Chinese. In the current study, we use modesty and politeness for kyemson and kongson when a translation is needed. We should, however, emphasize that the study is not focused on the translations of English modesty or politeness, but kyemson and kongson.

As briefly mentioned at the beginning of the paper, laypersons and researchers have presented rather mixed, cross-defined definitions for the two lexemes. For example, Hong (2020: 87) defined kongson (‘politeness’) practices as “the act of linguistic expression of modesty and propriety”. Cho (1980) claimed that the principle of kongson lies with the mind of kyemson. In addition, kongson was associated with lowering oneself, elevating others, and awareness of face (Kim 1997; Yang 2021) all of which were associated with kyemson, too (Kim 2015; Kim and Kim 2004). The indexical relationships between the two lexemes and Korean societal structures were also unclear. From a historical perspective, Lee (2002) argued that kyemson was a top-down moral requirement, whereas Kim and Kim (2004) empirically found that kyemson was manifested more in equal relationships. Kongson appeared to be less controversial in terms of its usage by inferiors to superiors. Nevertheless, the motivation for the inferior to do so was argued to be mitigating the psychological burden of the superior by Lee (1996), in contrast to Jung (2014) who considered that the inferiors had no choice.

Tao et al. (2016) presented one of the few metapragmatic studies on kongson in Korean and compared it to limao in Chinese and teinei in Japanese. The three are all Sino-words that have often been translated into politeness. However, they are not cognates and are written differently and used differently in Chinese. Tao et al. (2016) asked their L1 Korean participants to illustrate the social qualities that they imagined when hearing kongson. Based on the metapragmatic reports from 100 participants, they interpreted kongson as polite behavior, ways of speaking, courtesy, manners, being refined and cultivated, good image, good breeding, hierarchical relationship, personal relationship, and friendliness.

To the best of our knowledge, there have not been any metapragmatic studies of kyemson (‘modesty’). As a point of comparison, its Chinese cognates, qianxu and qianxun, were found to be a moral virtue and an important interpersonal tactic in Chinese (Bond et al. 2012; Xiong et al. 2018). Shi et al. (2021) categorized the social traits of a modest Chinese person into 34 indicators. The central ones include low-key, politeness, friendliness, humility, easygoing, sincerity, stability, and prudence. Gu (1990) and Zhou and Zhang (2018) proposed modesty as one of the core values of Chinese politeness and paired it with respectfulness. Zhou and Zhang (2018) argued that modesty in Chinese is practiced by self-effacement, in particular when the interlocutors “are more highly ranked on various types of hierarchical scales than others” (Haugh et al. 2015: 81). In short, there has not been any systematic understanding of kongson and kyemson not only in terms of their definitions, but also the contexts of their usage and their metapragmatics, rendering the current study a necessity.

In Korean, kyemson and kongson function as nouns, in contrast to their Chinese cognates being adjectives. They are compounded with -hata (‘to do’) to form adjectives kyemson-hata (‘be modest’) and kongson-hata (‘be polite’). -ta is the verb ending suffix, which can be replaced by other conjugations – for example, kyemson-han (‘modest’ as a modification), kongson-ha-key/kongson-hi (‘politely’). These lexical variations have been taken into consideration when extracting our data.

4 Methods^[1]

5 Findings

6 Discussion

The interdisciplinary methodologies used in this study have effectively differentiated the two bewildering politeness lexemes: kyemson and kongson. Specifically, kyemson is associated with personal qualities that one needs in managing a rapport relationship, regardless of their position in the structural hierarchy. The finding explains why kyemson was previously observed in both symmetrical and asymmetrical relationships (Kim 2015; Kim and Kim 2004). In contrast, kongson is related to a rather normalized set of embodied movements, including hand moves, bowing, and the way of speaking. They are sensitive to the social hierarchy and are primarily adopted by inferiors to superiors. This finding largely agrees with existing conclusions about the contexts where kongson should be used (Hong 2020; Kim 1997).

It was previously unclear which of kongson and kyemson was manifested by lowering oneself and elevating others (Kim 1997, 2015]; Kim and Kim 2004; Yang 2021). The current findings clarified the nuanced differences – that is, kyemson is associated with lowering oneself and elevating others at an abstract and largely attitudinal level, whereas kongson presents concrete practices of lowering oneself and elevating others, such as bowing. Both seem to align with Haugh’s (2004: 105) categorization of “self-oriented politeness” and “other-oriented politeness” wherein “one does not think too highly of oneself” and “one thinks well of others”, respectively. However, we may argue that the semiotically lowering and elevating practices of kongson (‘politeness’) appear to be more “politic” (Watts et al. 1992) than actually thinking well of others. As found in Section 5.3, inferiors practice kongson to their superiors, with little demonstration of the quality of their relationships. In this sense, kongson, translated into politeness, pertains to conventional politeness or discernment-driven politeness that observes the social status of a speaker instead of face needs (Hwang 1990; Ide 1989).

In terms of metapragmatics, one common aspect between kongson (‘politeness’) and politeness lexemes in English and Japanese is their association with appropriateness (Fukushima and Haugh 2014; Pizziconi 2007). The semiotic practices of kongson are deemed to be appropriate when addressing a hierarchical superior, as substantiated in Examples (6b) and (7a). If Lee (1996) was correct about the psychological motivation for practicing kongson – namely, to mitigate the superior’s psychological burden, kongson also shares the quality of “showing consideration for others” with English and Japanese politeness (Haugh 2004; Pizziconi 2007). However, the current study has not found any quantitative or qualitative indications of whether the consideration is genuine or not. The absence of an association with genuine other-oriented considerations, such as warmth, kindness and sincerity, may have differentiated kongson from English politeness and Japanese teinei (Haugh 2004; Pizziconi 2007).

In addition, our findings only partially agree with previous metapragmatic explanations for kongson (Tao et al. 2016). Tao et al. (2016) associated kongson with polite behavior, ways of speaking, courtesy, manners, being refined and cultivated, good image, good breeding, hierarchical relationship, personal relationship, and friendliness. Our findings support the inclusion of polite behaviors, ways of speaking, manners, and hierarchical relationships, and further reify the “polite behaviors” into multimodal movements, such as bowing and gathering hands. We do not support, however, the links between kongson (‘politeness’) and friendliness or personal relationships. The two are found in relation to kyemson (‘modesty’) in this study.

The difference between our findings and those of Tao et al. (2016) may stem from the size and types of data used.Tao et al. (2016: 139) employed written questionnaires, with which they asked 49 Korean males and 51 Korean females only one question: “What do you imagine at first when hearing the term kongson?”. Only five male participants (3.5 %) and nine females (9 %) mentioned friendliness, and four females (4 %) mentioned personal relationships in their answers. In contrast, the current study surveyed 9,718 texts where kongson (‘politeness’) was used. It is possible that a small portion of these data (e.g., 5 % as Tao et al. 2016 found) mentioned friendliness and personal relationships with kongson (‘politeness’). They are, however, not salient enough to become patterns that can be used to characterize the metapragmatics of this lexeme. The link between kongson and friendliness/personal relationships appears to be either individual variation in our data, or cognitively less salient in comparison to the other cooccurrences identified.

In terms of metapragmatics of kyemson (‘modesty’), the findings showed a consistency at a high, abstract level with its Chinese cognate qianxun that represents both a virtue used to evaluate one’s personality and behavior, and a tactic to manage social relationships (Bond et al. 2012; Xiong et al. 2018). In order to obtain or maintain a positive social evaluation, speakers with different social statuses were found to demonstrate that they have the virtue kyemson. In the meantime, kyemson was identified as an effective strategy to manage different types of social relationships, ranging from interpersonal to international (see Example 4). In contrast, kongson (‘politeness’), which is adopted to respect the social hierarchies, speaks little about the management of interpersonal relationships. Structural hierarchies have long been regarded as a feature of Korean social relationships (e.g., Kang 2003), against which Korean politeness is examined (e.g., Lee and Ramsey 2000; Yoon 2004). However, our findings point out a potential inconsistency between the two. Rapport relationships and social hierarchies may in fact be managed by different strategies, such as demonstration of kyemson and practices of kongson. Possibly, a violation of the kyemson norms indicates a speaker’s social status being powerful, and a violation of the kongson norms incurs relational damage (e.g., not being kongson to superiors) or demonstrates extra efforts to establish relational harmony (e.g., being kongson to inferiors). The interface between social hierarchies and social relationships in Korean is, nevertheless, more complicated than one featuring another.

Besides the high-level similarities, Korean and Chinese modesty lexemes – namely, kyemson and qianxun – differ significantly in the personal qualities that they are associated with. Kyemson co-occurs frequently with frankness and simplicity, which are not found in metapragmatic interpretations of qianxun (Shi et al. 2021). Low-key, stability, and prudence, which are associated with Chinese qianxun, are entirely absent with kyemson. The concepts of lowering oneself and respecting others occupied one of the topics of kyemson, showing their importance in understanding kyemson. They are, however, marginal in Chinese conceptualization of qianxun (Shi et al. 2021). These differences correspond to the observations of different societal changes in China and South Korea. While South Korea still values Confucianism as “a cultural grammar” (Kim 2016: 9), China abandoned Confucianism during the Cultural Revolution, untying the lowering/elevating practices from the norms of social interactions.

7 Conclusions

By implementing a computational technique, the current study unraveled the metapragmatics of two politeness lexemes, kyemson and kongson, which are otherwise hard to explain from both emic and etic perspectives. It contributes to the increasing volume of studies on metapragmatics of politeness, by providing a novel approach that can help researchers overcome a number of challenges, including the analysis of broader co-texts, the involvement of diverse contexts, the exploration of politeness-in-use, the multiplicity of metapragmatic interpretations, and the extraction of regularities from discursivities. The findings also provide insight into the underexplored Korean politeness lexemes by revealing their associations with different social qualities and practices. They can be used to define the two lexemes, explain the contexts of their usage, and assist non-Korean speakers in understanding their connections to Korean social cultures.

We should note that there are many more politeness lexemes, such as yeyuy in Korean (Kim and Brown 2024) and reigi, teinei, and keii in Japanese (Haugh 2004), as well as Sino-cognates of kyemson and kongson in other Asian languages (e.g., Vietnamese), that need to be disentangled for the ease of understanding by their global audiences. Therefore, we recommend the application of our methods in future examinations of intracultural and cross-cultural metapragmatic variations of politeness lexemes.

We should also offer several caveats for the application of our methods. First, although the current study adopted a relatively large size of data from both spoken and written corpora, the spoken data were transcribed with inevitable information loss of their prosodic and non-verbal features. It will be a welcome enrichment if future data analysis includes both transcribed spoken data and their original recordings. Second, topic modeling techniques use vector spaces to extract patterns. If the vector space is highly sparse, for example, including a less frequent politeness lexeme or when its co-texts only occasionally cluster together, other methods, especially manual text analysis, may produce more reliable results than topic modeling techniques. Finally, we should note that BERTopic represents only one of the effective computational methods that can be used to explore politeness metapragmatics. Future studies are encouraged to explore other methods, such as word embeddings and other machine learning techniques, that serve the purpose of their research.