Compliment responses in Hong Kong: an application of Leech’s pragmatics of politeness


 This paper focusses on compliment responses in the context of group undergraduate student projects, in a university in Hong Kong. It applies Geoffrey Leech’s model of politeness. Although less well known than some other politeness models, it has proved relatively popular in studies of compliment responses, which are often said to involve a clash between the Maxim of Agreement (achieved by agreeing with the complimenter) and the Maxim of Modesty (achieved by mitigating the compliment, thereby disagreeing with the complimenter). This study deploys Leech’s most recent work on politeness in the study of compliment responses. Using an innovative variant of the discourse completion task adapted to WhatsApp to collect text messages and metapragmatic comments from undergraduate students in Hong Kong on their messages, it reveals that acceptance strategies are overwhelmingly the most frequent type. This finding adds to the small body of work on compliment responses in Hong Kong cultures, and, more generally, to cross-cultural pragmatics studies on compliment responses. However, the interpretation of this result needs to attend to the detail. The key specific acceptance strategy in our data is the expression of gratitude, and this, we argue, is best accounted for through the Maxim of Obligation, a maxim proposed in Leech’s more recent work.


Introduction
This paper investigates the compliment responses (CRs) of Hong Kong students. CR production by Hong Kong Chinese seems particularly interesting to study, given more detailed account. According to Leech (1983: 82), the Politeness Principle (PP) as a corrective to the concentration of recent studies focussing on the "bigpicture view of how politeness relates to social behaviour and society in general" (2014: ix). Leech (2014: 15) also reflects on criticism of his terms "absolute" and "relative" politeness from his earlier work. He renames these terms pragmalinguistic and sociopragmatic politeness, respectively, to better reflect his stance that politeness can be studied from a more linguistic perspective (pragmalinguistic) and from a more societal or situated perspective (sociopragmatic). These are notions that we incorporate into our methods for a larger study, of which this is a part. In this paper, like Leech, our focus is mostly on the pragmalinguistic side.
In Leech (2014), the PP is defined as a "constraint observed in human communicative behaviour, influencing us to avoid communicative discord or offence, and maintain or enhance communicative concord or comity" (p. 87). Moreover, he sees the General Strategy of Politeness (GSP) as a super-constraint, and defines it thus: "In order to be polite, S expresses or implies meanings that associate a favourable value with what pertains to O or associates an unfavourable value with what pertains to S" (S = self, speaker; O = typically the addressee, but could be a third party) (2014: 90). 2 The maxims are considered more specific realisations of the GSP and are couched in terms of giving a high or low value to the wants, qualities, obligations, opinions or feelings of S or O (Leech 2014: 91-98). The maxims are increased to 10, and are labelled thus: Generosity, Tact, Approbation, Modesty, Obligation (of Speaker/Self to the Other), Obligation (of the Other to the Self/Speaker), Agreement, Opinion reticence, Sympathy and Feeling reticence.
One criticism of Leech's 2014 book is that it dwells almost entirely on English data, and builds theory on them (Terkourafi 2015: 958). Leech had begun to address this criticism in his 2007 paper, which draws examples particularly from Chinese, Korean and Japanese to illuminate the proposed east-west divide in politeness, and Brookins (2010), for example, had applied Leech's framework to Latin and Ancient Rome. In Klégr's (2016: 72) view, Leech demonstrates in his 2014 book that his revamped model is "capable" of handling other languages. However, as Leech (2014: x) concedes, his new book still focuses primarily on English and, in fact, on only two varieties, American and British, with a stronger slant towards British English. We see it as a logical next step to test his model on another type of English and culture. Moreover, Hong Kong's cultural hybridity, with influences from colonial Britain and traditional China (Bolton 2002;Lai 2011), may lead to some interesting findings on CRs and politeness.

Compliment responses
To better understand CRs, a brief description of its adjacency pair is helpful. Holmes (1986: 485) defines a compliment as "a speech act which explicitly or implicitly attributes credit to someone other than the speaker, usually the person addressed, for some 'good' (possession, characteristic, skill, etc.) which is positively valued by the speaker and the hearer". Ruhi (2006: 47) argues that a compliment carries both an assertive and an expressive force and thus affords the recipient latitude in how to respond. According to Pomerantz's (1978: 83) seminal work on CRs, recipients can respond by accepting/rejecting the expressive force or agreeing/disagreeing with the underlying assertion. This idea is also supplemented by the claim that agreeing is associated with acceptance and disagreeing is associated with rejection. Leech's (2014: 189) Modesty maxim, along with the Agreement maxim, explicitly surfaces when the recipient chooses to agree/ disagree with the assertion. Leech argues that compliment recipients are placed in the double bind of having to choose between the two maxims. In addition to agreeing/disagreeing or accepting/rejecting, a recipient can also respond in ways that indicate deflection/evasion (Ruhi 2006). These competing constraints and varied response options aside from (dis)agreement have made CRs interesting to study, particularly in English and Chinese cultures.
Studies of CRs in Englishes, whilst noting some variation, point to a general preference for accepting compliments (e.g. Chen and Yang 2010;Herbert 1989;Holmes 1986;Miles 1994;Rose 2001). Explanations as to this finding differ. While Holmes (1986) explains CR behaviour as influenced by the tension between Leech's Agreement and Modesty maxims, Herbert (1989) argues that agreement and self-praise avoidance are likely culprits. Turning to Mainland Chinese CRs, an early study comparing American English and Chinese CRs is Chen (1993). He found large differences in the frequency of acceptance and rejection strategies. The American English CRs showed acceptance at 39% and rejection at 13%, with the remaining CRs exhibiting compliment returns and deflections. The Chinese CRs, however, showed acceptance at just 1% and rejection at 96% (with the remaining CRs exhibiting a thanking plus denigration strategy)a result which suggests the key importance of modesty. This led Chen (1993) to claim that, while Brown and Levinson's politeness theory could account for the American English CRs, Leech's politeness theory could account for both the Chinese and English CRs, primarily through the Agreement and Modesty maxims. Similar findings were reported by Tang and Zhang (2009) investigating Australian English and Chinese CRs: Chinese CRs exhibited a higher preference than the English CRs for evasion and rejection.
They also explained these differences in reference to Chinese norms of modesty through self-denigration.
In another study (Cheng 2011), support for this explanation can be found in metacomments from retrospective interviews of Chinese ESL students in America on their English CR production. A number of these comments point to their understanding of English CRs as requiring thanking, agreement, and certain evasion strategies that avoid rejecting the CR to maintain rapport, despite also reporting that they would normally not agree with CRs and sometimes reject CRs in Chinese (Cheng 2011(Cheng : 2210(Cheng -2211. There is yet further support in quantitative studies that investigate cross-linguistic and cross-cultural influence on CR production. In a study of Chinese-English bilinguals producing CRs on Facebook and Renren, 3 Eslami and Yang (2018: 18) found considerable differences in acceptance, evasion and rejection between Chinese and English CRs, leading the authors to posit that "when using Chinese on Renren (a national SNS) participants were inclined to activate Chinese cultural schema of modesty, and when using English; they were inclined to activate Western cultural schema of agreement". A similar pattern of more acceptance and less rejection in English CRs than in Taiwan Chinese CRs was found in Yu (2004: 107), although the difference was less profound. Of course, one cannot claim that all varieties of English and Chinese are homogenous in CR behaviour. Yuan (2002), for example, concludes that CRs in Kunming Chinese exhibit more acceptance and much less rejection than the CRs in Mandarin Chinese from Chen's (1993) study.
Regarding Hong Kong CR production, Rose (2000: 52-54) observes that Cantonese and English CR production by Hong Kong primary school students favoured acceptance over deflection and rejection. 4 A separate but perhaps more interesting finding is that the Cantonese CRs exhibited considerably more rejection and considerably less acceptance than the English CRs did, although acceptance remained the favourite. This is consistent with Rose and Ng's (2001: 165) finding that English CRs by Hong Kong undergraduate students prior to teaching instruction exhibited much higher frequencies of acceptance than deflection or rejection. Spencer-Oatey et al. (2008) studied the evaluations of CRs by the British, Mainland Chinese and Hong Kong Chinese respondents according to appropriateness, conceit 5 and impression (favourable/bad). Conceit obviously relates to 3 Renren is Chinese social networking site. 4 It is also worth noting that the researcher was informally told by teachers that accepting English CRs was taught to students. 5 We assume the sense of proud or arrogant here. We wonder what informants made of it, as it is a rare word in English, a mere 96 appear in the 100 million-word British National Corpus, some of which refer to a literary device. the degree of modesty. The acceptance strategy again received the most favourable scores among all cultural groups, although the Hong Kong Chinese respondents were, surprisingly, the most positive about rejection and disagreement strategies among all cultural groups. This, one might think, is odd since Hong Kong, as an international city and a former British colony, should have more Western influence and thus one would expect their evaluations of rejection and disagreement to be less positive than those of Mainland Chinese respondents. More studies of CRs in Hong Kong are needed.
Based on the studies discussed, it seems that Leech's Agreement and Modesty maxims play a key role in accounting for English and Chinese CR behaviour. However, analysing Turkish CRs, Ruhi (2006) makes an important criticism of the Agreement and Modesty Maxims in accounting for CRs. She makes the point that accounting for CR behaviour using only agreement and modesty is schematic and that other maxims may be more influential. One of our aims is to shed more light on Leech's maxims, especially those from his (2014) revised politeness model.

Data and methods
The data used for this study was taken from a larger study (Pat forthcoming) on politeness in the following context: WhatsApp group chats amongst university undergraduates doing group projects in Hong Kong. These group chats generally involve English, although they can include Chinese characters or Cantonese converted into roman alphabets.
Our data was collected through DCTs that were given an original twist by adapting them to WhatsApp, a digital communication medium that has over 1 billion users in over 180 countries (About WhatsApp 2019) but has largely been neglected by politeness scholars. By adapting the DCT to simulate the conditions of typing on WhatsApp, much of the criticism relating to the provision of written responses to DCT scenarios that are clearly spoken scenarios is addressed (Economidou-Kogetsidis 2013: 24). Our DCT deploys a medium, i.e., typing a message on a smartphone, that would be a natural choice for the students enacting the situational scenarios we constructed for them. Another frequent criticism is that DCTs have situational scenarios that do not include adequate contextual information (Billmyer and Varghese 2000;Woodfield 2008: 63). To address this criticism, we converted Leech's (2014) notions of pragmalinguistic politeness and sociopragmatic politeness into methodological tools to elicit messages with CRs, specifically a pragmalinguistic DCT and a sociopragmatic DCT, with this paper being more focussed on the former. Similar to other CR studies (e.g. Chen 1993; Chen and Yang 2010; Tang and Zhang 2009), the pragmalinguistic DCT provides a CR situational scenario with limited contextual information. The sociopragmatic DCT provides the same limited information but with additional contextual information, which attends to the aforementioned criticism. Nevertheless, we are under no illusion that our adapted DCT can collect truly authentic data, but it does create a closer approximation to how CRs would be produced in real WhatsApp group chats and is effective for understanding realisation patterns at the initial research stage (Beebe and Cummings 1995: 80;Yuan 2001: 289).
Furthermore, some CR studies deploying DCTs instruct participants to produce CRs that are socially appropriate or most natural in a particular scenario (e.g. Chen 1993; Chen and Yang 2010), and from there researchers extrapolate about politeness. In our pragmalinguistic DCT, participants were asked to produce CRs with either a high level of politeness or a low level of politeness. Theoretically, it is likely that a low level of politeness pertains to CRs that are socially positive, appropriate and expected, whereas a high level of politeness pertains to CRs that do more than this (Locher and Watts 2005). By allowing participants to demonstrate their own understandings of the concept that the label "politeness" evokes and how it applies to the way CRs are formed, we avoid relying on concepts determined and applied by researchers, an approach which has been critically discussed (Eelen 2001;Watts 2003). Comparing CRs produced from each of the two politeness levels allowed us to explore a possible politeness dimension. We, nevertheless, also collected CRs that are appropriate or most natural in a particular scenario from our sociopragmatic DCT. Lastly, we deployed Google Docs to collect participantdriven metacomments about their own CR messages, which we will use to supplement our analysis of the CRs.

DCT design
The pragmalinguistic DCT was designed to elicit the range of CRs that participants might generally use in WhatsApp group chats during group projects. It includes general information about the context and instructs participants to "respond to compliments from the group on your PPT design" in two ways: the first exhibiting a high level of politeness and the second exhibiting a low level of politeness (Appendices A-Cwhich can be accessed as Supplementary Material from the journal's webpagesprovide further information through screenshots of the various types of DCTs we deployed and the Google Docs collection of metapragmatic comments). There are also rating scale questions, multiple choice questions, and a background survey but they are not a part of the current study.
There are two versions of this pragmalinguistic DCT. The first version was administered in classrooms and required participants to use their mobile phones to complete the DCT through a Qualtrics online questionnaire. This version was not administered on the actual WhatsApp mobile application due to technological difficulties with collecting the data and ethical concerns with collecting mobile numbers that could not be deleted in the presence of the participants as a result of the large number of participants in each classroom. Instead, participants were instructed to imagine that they were a part of a group project and that they were typing into a WhatsApp chat. This first version was anonymous. The second version was conducted one-on-one or in small groups with the second author of this study. CRs were produced on WhatsApp and sent via WhatsApp to the second author. Once downloaded, participants' mobile numbers were deleted in the presence of the participants. The CRs were then immediately transferred to a Google Doc for participants to comment on, which is explained in the next subsection.
The purpose of the sociopragmatic DCT was to elicit CRs that participants might use on a specific occasion of use in a more contextualised WhatsApp group chat. It provided much more information regarding the particular instantiation of the group project, including the background of each groupmate and their social distances amongst each other. More information on the CR context such as the semantic content of the compliment and the physical location of the participant during the mobile typing of the CR was also provided. This DCT was administered through a Qualtrics online questionnaire in classrooms to a subset of the participants who also completed the pragmalinguistic DCT. It was administered approximately two weeks after the pragmalinguistic DCT to reduce the chances of participants remembering the details of that DCT. This DCT also included a background survey and was anonymous.

Metapragmatic comments via Google Docs
The participants who completed the pragmalinguistic DCT one-on-one or in small groups with the second author (the second version mentioned above) were also invited to provide metapragmatic comments on why their CRs were polite. To reduce researcher influence on what the participants commented on, the second author transferred the participants' CR data onto a Google Doc, which instructed participants to use their computers and the Google Doc comment function to indicate which parts of their CRs were polite and why they were polite. While the CRs from the sociopragmatic DCT enlarged our sample of CR strategies, metapragmatic comments were not collected for the CRs produced from the sociopragmatic DCT.

Participants
The participants were undergraduate business major students at The University of Hong Kong in the Spring of 2017. Only students who identified their home city as Hong Kong, spoke Cantonese with family or friends and did not spend a substantial amount of time living in an English-speaking country were included. This was to ensure that the participants were of Hong Kong Chinese ethnicity and were not international students or mainland China students who were temporarily studying in Hong Kong. One hundred and sixty-four students who completed the pragmalinguistic DCT matched this profile, and 93 who completed the sociopragmatic DCT matched this profile. Although there were no measurements of the students' English capabilities at the university, the students participating in this study had IELTS-equivalent scores of above 6.31. 6 Lastly, ethics approval was attained from the University of Hong Kong and Lancaster University. The participants who completed anonymous questionnaires were given an information sheet while informed consent was sought from those who provided data in the presence of the second author.

Methods of analysis
Coding of the CRs given in the DCTs followed the acceptance, deflection, and rejection scheme from Holmes (1986). We adapted sub-strategies of this scheme devised by Ruhi (2006) and Chen and Yang (2010) and added new sub-strategies where necessary in creating our coding manual (see Appendix D, which is downloadable from the journal's webpages). The initial coding of the CRs was done by the second author of this study. The first author then conducted an interrater reliability check of the coding with a random subset of 20% of the CRs from the pragmalinguistic DCT. There was an approximately 85% agreement in the strategies coded for this subset. Afterwards, the initial coding was rechecked.
Regarding the metapragmatic comments relating to the pragmalinguistic DCT, like Cheng (2011), we identified the major themes and the sub-strategies that they were typically referring to. These comments were analysed as one group rather than as two separate groups (i.e., comments on high politeness CRs and comments 6 These figures are no longer officially published on the university website, although they were accessed and recorded by the second author of this study when they were available. There were two acceptable admissions English exams: the HKDSE in Hong Kong or IELTS. The minimum score achieved by the participants of this study who took the HKDSE was 6.31 as benchmarked by Hong Kong Examinations and Assessment Authority and the minimum entrance requirement for those who took IELTS was 6.50. on low politeness CRs). This allowed for a higher number of comments for each major theme to be considered, which more convincingly established each major theme. As the participants chose and indicated the parts of their CR message that they were commenting on, the individual strategies of the parts of each message that received comments were also tallied.

Compliment responses: findings
In this section, the findings will be divided and presented in three parts: (1) general patterns of all CRs; (2) patterns between the high and low level of politeness CRs from the pragmalinguistic DCT and (3) metapragmatic comments on CRs. As the scope of this paper is on the pragmalinguistic side, the data from the sociopragmatic DCT will not be discussed but tabulated as part of the general patterns for a larger sample of CRs. Examples for Sections 4.1 and 4.2 can be found in the CR coding manual in Appendix D (see Supplementary Materials on the journal's webpages).

General patterns of all CR messages
A total of 364 valid CRs were produced from 140 pragmalinguistic DCT questionnaires and 84 sociopragmatic DCT questionnaires. 7 Of the total 364 CRs, 80 were produced under conditions where the second author was present with either one participant or a small group of participants, which did not exceed four at any time. The others were produced under large classroom settings. The majority of the CRs included the use of more than one CR strategy. Each strategy was coded into the categories as stated and exemplified in the CR coding manual. Table 1 below presents the frequencies of each strategy, which are further grouped according to the superordinate strategies of acceptance, deflection and rejection. These frequencies are given according to each of the two types of DCTs, although only the combined frequencies will be discussed. Table 1 shows a distinct preference (64%) for acceptance strategies in CRs, followed by deflection strategies (27%), with rejection strategies dispreferred by a wide margin (8%). These macro pattern results are nearly identical to those found in Chen and Yang (2010) for CRs in Mandarin Chinese and similar to those found in Rose and Ng (2001) for English CRs in Hong Kong. The most preferred specific CR acceptance strategy was expressing gratitude, followed by expressing gladness, and then the deflection laughter strategy. These results suggest that rejection of CRs is not nearly as preferred as once thought in Chinese cultures (see Chen 1993), which Chen andYang (2010: 1959) tentatively attribute to more recent Western cultural influence.

Patterns between the high and low levels of politeness CRs
In Section 3, it was mentioned that the pragmalinguistic DCT provided minimal information on the context of the compliment in order to elicit the participants' knowledge of the CRs they would generally produce in this kind of context. The DCT asked the participants to produce CRs they associated with a high level of politeness and CRs with a low level of politeness, so that the difference in strategy frequency between the two levels might indicate what the participants believed to be strategies that enhance or reduce politeness, leading to stronger claims that a particular strategy is polite. This could also tell us what specific strategies the participants are likely to enact in occasions requiring a high level of politeness and in occasions requiring a low level of politeness, although the exact occasions will clearly require further research. Table 2 shows the frequencies of each sub-strategy from the high politeness CRs and low politeness CRs under the same superordinate categories used in Section 4.1. The CRs follow the same general pattern from the macro view in the previous section. Again, acceptance strategies are the overwhelming favourite followed by deflection and rejection. Surprisingly, however, the percentage differences between high and low politeness CRs for these superordinate strategies appear very small, with each of the three shifting in frequency by less than 2%. In other words, the proportions of Acceptance, Deflection and Rejection do not change much when reducing the level of politeness. In terms of percentage differences, the specific strategies expressing gratitude and expressing gladness increase and decrease the most, respectively, though the differences are modest. These changes may hint that expressing gratitude in CRs is conventionally polite in this context (see also Cheng 2011Cheng : 2210Herbert 1989: 5-6), whilst expressing gladness in CRs is a preferred option for enhancing politeness. More research on specific strategies is clearly needed here.
Nevertheless, viewing the overall trends, the raw frequency of each superordinate strategy is much reduced, going from high to low politeness CRs. In fact, almost all sub-strategies decrease in raw frequencies. Thus, a major general difference between more politeness or less politeness is simply the quantity of strategies used. It is also interesting to note the large raw frequency increases and decreases in some specific sub-strategies. The raw frequency drops in expressing gratitude and expressing gladness are the largest, and this likely signals their importance in producing polite CRs. Conversely, praise upgrade may not be polite as an acceptance strategy because it increases from zero occurrences for high politeness to nine for low politeness. Additionally, the considerable drop in wellwishing, informative comments, share credit, denigrating, and reciprocation could potentially indicate that these are amongst the politer strategies.

Metapragmatic comments on CRs
This section presents the findings from the metapragmatic comments given by 40 participants who completed the pragmalinguistic DCT in one-on-one or small groups settings with the second author of this study. Table 3  themes that were found in the comments, the number of comments that referenced each theme, and the frequency of each strategy that was commented on for each theme. The remaining subsections analyse the comments and provide examples, along with the corresponding CR, relating to each theme.

Expressing appreciation or thanking as intrinsically polite
A preliminary point to make is that our data suggest people think of politeness as being in particular words, expressions or structures, as can be seen in examples [1]-[6] below. Some politeness scholars (e.g . Fraser 1990;Watts 2003) have argued that politeness is not intrinsic to the language but arises from a judgement of that language in context. Of course, the participants' focus on politeness in language is partly a function of our method, but it does suggest a readiness to see politeness as, in some way, intrinsic to some expressions. Almost half of the comments on why strategies were polite related to the expression of appreciation or thanking for the compliment. This is perhaps not a surprising finding since "one is […] supposed to say thank you to acknowledge a compliment" (Aijmer 1996: 70); in other words, at least in some cultures, we are socialised into response routines of appreciation or thanks. In [1], the positive evaluation in the compliment essentially obligates the complimentee to reply with appreciation or gratitude. Comment [3] especially indicates that expressing  Appendix C. These boxes often had more than one comment. Additionally, one comment could refer to multiple strategies in a CR. The number of comments for each major theme does not total  as a number of comments could be categorised into multiple themes. b The surprising finding that the praise upgrade strategy is in the humbleness theme is due to a participant commenting on the strategy as "arrogant".
gratitude, especially in terms of thanking, is a bare minimum for CRs. It is interesting to note that [6], combined with the fact that the comment was given for a disagreement strategy from a low politeness CR, implies that expressing gratitude or appreciation is more important for politeness than disagreeing is. This suggests that if modesty is a key influencing factor in producing polite CRs, disagreement may not be an optimal strategy towards expressing modesty. Some comments, such as [4] and [5], relate to the sincerity of the gratitude expression; the politeness of the gratitude or appreciation expression is deemed to be subject to the speaker's sincerity (see also example [7]).
[1] A way to thank people for saying that my work is good. This is polite when people make compliments.
[2] polite word [3] This is the basic politeness by saying thank you [4] a complex sentence structure just seemed to make my gratitude more sincere [5] one word of "Thanks" lowers level of grateful expression. It may make others doubt if I am truly thankful or not.
[6] This sentence is not very polite such that I do not appreciate their compliments.

Politeness strategy modification
The comments under this theme show how the participants were cognisant of the potential to intensify the politeness of CR strategies through modification. Comment [7] illustrates how the participants were aware of the lexical modifier 'so much' to enhance the sincerity of the politeness in thanking, which is similar to comment [5] above. In [8] below, it appears that the syntactical complexity of a CR also plays a role in modifying the sincerity of the politeness, although the message is not really complex and just includes a comma separating two iterations of the expressing gratitude strategy. [7] CR message 8 : Thank you so much. Everyone did a great job in this project. Thanks Participant comment: Use 'so much' to make it sounds more sincere [8] CR message: Thanks for the compliment, really appreciate that:pp Participant comment: a complex sentence structure just seemed to make my gratitude more sincere The participants were also aware of text messaging resources for enhancing politeness such as letter repetition and emojis, as seen in comment [9] below. Recent research on text messaging has shown how users can exploit the text messaging medium for politeness purposes (e.g. Darics 2010;McSweeney 2018).
Interestingly, it appears so easy to intensify politeness pragmalinguistically in [9] that the participant feels it is actually less polite due to potential insincerity. Most of the comments on modification as a politeness resource were given for the expressing gratitude strategy, which is an easy strategy to intensify but difficult to mitigate due to its politeness potential (Holmes, cited in Wong 2010).
[9] CR message: yeah thanks so muchhhhh!! we all contributed to it;) Participant comment: with the repeated "h" and the smiling face, I hope to express my great gratitude to my group mates, though this may be less polite as it seems exaggerating and not a whole-hearted message

Humbleness
The third most frequently recurring theme is humbleness and arrogance avoidance as motivating factors for CRs. Clearly, these are concepts that are directly related to or subsumed by Leech's (2014) Modesty maxim. The comments directly reference the concepts of humbleness or arrogance when describing the CRs, as in [10] and [11], or imply them, as in [12].
[10] CR message: thanks haha just picked this simple theme lol seems that it suits our presentation Participant comment: yea trying to be humble; trying to be humble? 9 [11] CR message: Thanks. It took me whole night to finish it. Good to hear that. Participant comment: Try not to be arrogant [12] CR message: Ngawww you are welcome!! Glad that you guys all liked it! Let's continue to work hard for our presentation laterrr!!! Ooshhh we need to get As!! Participant comment: divert attention from praising me Interestingly, these modesty concepts do not appear with only deflection and rejection strategies. This finding differs from previous studies suggesting that deflection and rejection strategies are an indication of modesty (e.g. Chen 1993;Eslami and Yang 2018;Holmes 1986;Tang and Zhang 2009) and implying that this is not the case for acceptance strategies. While the participants also viewed deflection and rejection strategies as humble in our study, their modesty comments given for acceptance strategies such as expressing gratitude and expressing gladness, as shown in [13] and [14], demonstrate that modesty needs to be conceived more broadly in its operation in CRs. This is especially worth exploring in research situated in Chinese contexts where the stereotype of complimentees preferring to deny CRs and avoid expressing gratitude remains (see Wong 2010Wong : 1256

Positive emotions as politeness
The expression of positive emotions was also an aspect of politeness. This ties in with researchers who have alluded to positive emotion being a key feature of politeness (e.g. Arndt and Janney 1987;Locher and Watts 2005;Watts 2003 Additionally, while CR strategies such as using humour, laughter and expressing gratitude are easy candidates for the expression of positive emotion, expressing unease was also thought of as expressing positive emotion, as seen in [17]. One might expect unease to be a negative emotion, but its expression through an emoji converts the negative emotion into a playful one that attempts to build social closeness (McSweeney 2018: 153; Sampietro 2019).
[17] CR message: Thanks!!! Participant comment: three exclamation marks show sincerity and the emoji is cute enough.

Discussion and conclusions
In Leech (1983Leech ( : 137, 2014, the Agreement and Modesty maxims are posited as regulative principles for CRs. However, if these are the key explanatory constraints, it is equally clear from the literature reviewed in Section 2.2 that their application is culturally variable. Our findings suggest a clear preference for Agreement over Modesty: acceptance strategies accounted for 64% of the strategies and rejection only 8%. However, this conclusion only holds if one assumes that acceptance implies agreement and modesty is restricted to what is counted as a rejection strategy. 10 Our metapragmatic comments suggest that the participants do not restrict modesty to only rejection strategies. While there were metapragmatic comments related to the Modesty maxim, there were no themes related to the Agreement maxim. Moreover, modesty appears to lurk in the shadows of the CR strategies our informants use that scholars do not typically relate to modesty such as expressing gratitude and expressing gladness (see Table 3 and Section 4.3.3).
Our argument requires us to account for the particular acceptance strategies that are used to respond to a compliment. The strategy expressing gratitude, which includes appreciation, occurs with overwhelming frequencymore than double any other strategy (cf. Table 1). This is strongly supported by the participants' metapragmatic comments, where comments on the expression of gratitude outweigh any other comment by far (cf. Table 3). Our findings also suggested, contrary to some politeness scholars, that the participants were particularly attuned to the expression of formal aspects of politeness, that is, particular words and expressions (e.g. thank you). The danger here is that simply doing what is formal and conventional can run the risk of being deemed not enough, and hence we see a number of metapragmatic comments relating positive emotions to the intensification of conventional gratitude expressions. How is expressing gratitude accounted for in Leech's maxims? We would argue that neither Modesty, which has some support in our findings, nor Agreement are the most obvious maxims here. Instead, it is readily accounted for by the Obligation maxim in Leech's (2014) scheme. If a compliment is a verbal gift (Kerbrat-Orecchioni, cited in Mustapha 2011), then thanking may be a polite form of symbolic compensation (Haverkate 1988: 391). What regulates the expression of that compensation is the Obligation maxim; Leech (2014: 96) explicitly proposes that the Obligation maxim relates to the "expression of gratitude for some favour H has done to S".
We would go further and say that the basis of that obligation lies in (im) politeness reciprocity: "a constraint on human interaction such that there is pressure to match the perceived or anticipated (im)politeness of other participants, thereby maintaining a balance of payments" (Culpeper and Tantucci 2021: 150). A polite compliment pressures or obliges the addressee to reciprocate with polite gratitude, and the use of intensification ensures an appropriate balance of payments (though too much intensification risks a perception of insincerity, cf. Section 4.3.2). Furthermore, gratitude, whilst implying acceptance of the compliment, can have modesty built into it, unlike some other acceptance strategies (e.g. agreeing and praise upgrade in Table 1). Brown and Levinson (1987: 67) suggest that, in expressing thanks, a speaker "accepts a debt, humbles his own face". One consequence of our findings and account of them is that we have indicated that multiple maxims can be involved simultaneously in a single CR. Participants achieve a high-wire balancing act amongst several maxims, though the force of one or more maxims may be stronger on some occasions. We would argue with respect to expressing gratitude that the Obligation maxim exerts the strongest force.
Whether the Obligation maxim has the same weight in other cultures, and indeed specific contexts, is an open question. Studies on CRs contributing to crosscultural pragmatics have had somewhat mixed findings and narrow explanations, particularly with respect to Chinese cultures (see Sections 2.2 and 4.1). This could be because they were examining different Chinese cultures from different time periods, they had different analytical methods, or the researchers were overenthusiastically interpreting their results in terms of modesty whilst ignoring other possible notions. Our findings of acceptance strategies as most favoured in our specific context based in Hong Kong are consistent with other studies of other contexts based in Hong Kong. The fact that we find gratitude and appreciation to be so important is also consistent with studies of a number of other cultures (e.g. Chen and Yang 2010;Herbert 1989;Rose 2000;Ruhi 2006;Tang and Zhang 2009), but the notion of obligation lacks concerted research attention.
In conclusion, this paper has made three kinds of contribution: descriptive, methodological and theoretical. It casts additional light on the use of CRs and matters of politeness perception in the context of Hong Kong, noting, for example, the dominance of acceptance strategies. We took the traditional DCT and blended it with WhatsApp, a medium that members of the group projects in our study would normally use to communicate with other members. In doing this, we circumvented, to some extent, the typical problem of many pragmatics DCT studies, namely, eliciting data in one medium (writing) about scenarios that take place in another medium (speech). This paper gives more exposure to Leech's (1983) politeness model, which has generally been eclipsed by the more widely used Brown and Levinson (1987), and is perhaps the first to consider compliment responses (CRs) in full knowledge of both his 1983 work and his 2014 book-length update The Pragmatics of Politeness. We have argued that the Obligation Maxim, present in his 2014 work, along with the other maxims, offers a new and more adequate means of accounting for our findings.