1. Introduction¹

1.1. Politeness in computer-mediated communication

The aim of this study is to shed light on the use of argumentative concession in asynchronous online discussion forums. The use of online discussion forums and other types of computer-mediated communication (henceforth CMC) in educational settings has become an extended practice that enables participants to work and construct knowledge beyond the time and space constraints of the classroom. In fact, online interaction environments have been reported to be potentially powerful tools for collaborative learning and group communication (Schallert et al. 2009; Van Nguyen 2010). As predicted by Jordan et al. (2014: 451), CMC has continued to play “a significant role in formal learning as institutions of higher education increasingly offer online and hybrid courses,” especially with the challenges brought about by the COVID-19 crisis.

Against this background, we also seek to explore how learners participating in the discussions use concession in combination with other politeness strategies in a collaborative pedagogical context. Therefore, we are interested in issues of face (Goffman 1967) and politeness (Brown and Levinson 1978; 1987). In the list of potentially face-threatening acts (FTAs), Brown and Levinson’s theory of politeness includes orders, requests, suggestions, advice, reminders, warnings, offers, promises or criticism (Brown and Levinson 1987: 66-67). These speech acts can be mitigated by using positive and negative politeness strategies, depending on whether they are used to protect positive face (i.e., the universal desire to be appreciated and socially accepted) or to protect negative face (i.e., people’s desire to preserve autonomy). Examples of positive politeness strategies include attending to the interlocutor’s needs or wants, seeking agreement, softening disagreement, including the writer and the reader in the activity, and showing praise or appreciation, among others. Negative strategies, on the contrary, include being indirect, minimising an imposition, apologising, and impersonalising a situation, among others (Schallert et al. 2009: 718).

Even though Brown and Levinson’s work has remained influential over the years, it has been frequently challenged. Thus, considerable criticism has come from Watts (1992, 2003), Locher (2004), Locher and Watts (2005, 2008), who argue that Brown and Levinson’s model is not “in fact a theory of politeness but rather a theory of facework” that fails to account for “those situations in which face-threat mitigation is not a priority,” such as aggressive or impolite behaviour (Locher and Watts 2005: 10). Focusing on the interpersonal dimensions of language used in interaction, they develop the concept of ‘relational work’, i.e., “the ‘work’ that individuals invest in negotiating relationships with others” (Locher and Watts 2005: 10). It is important to remark that, in their view, Brown and Levinson’s concept of politeness can still be used, but it should be viewed as only a small part of relational work, which, in turn “comprises the entire continuum of verbal behaviour from direct, impolite, rude or aggressive interaction through to polite interaction” (Locher and Watts 2005: 11).

From the point of view of politeness, the online medium has several peculiarities which inevitably shape CMC interactions. On the one hand, it imposes certain limitations which make participants reinforce the interpersonal links with their partners using markers of affection, interactive responses, and group cohesion expressions (Fernández-Polo and Cal-Varela 2017). On the other hand, the lack of non-verbal clues increases the importance of using politeness to avoid misunderstandings, since FTAs such as “disagreements, criticisms, requests for information or help, and requests for clarification of a prior message” (Schallert et al. 2009: 715) are typical of CMC interactions (Herring 2023). This is especially true for those interactions including assessment or evaluation of peers’ (L2) writing (Cal-Varela and Fernández-Polo 2019; Pyo and Lee 2019), as is the case with the discussion forums in this study (cf. section 2). In these language learning contexts, where the emerging virtual communities have been found to promote interaction and diminish anxiety of communication (Deris et al. 2015: 79), the presence of FTAs also leads participants to soften their comments through mitigation strategies.

The emerging interest in politeness issues in CMC has produced a substantial body of research. Different CMC modes have been covered in the literature: e-mails (Harrison 2000; Vinagre 2008), Wiki exchanges (Li 2012), blogs (Puschmann 2010), and synchronous and asynchronous discussion forums (Herring 1994; Park 2008; Schallert et al. 2009), among others. In general, positive politeness strategies have been found to be more frequent than negative strategies in CMC. This is often attributed to the participants’ need to create solidarity (Park 2008; Vinagre 2008) and to maintain accuracy, while avoiding the ambiguity and indirectness that is often brought about by negative politeness (Morand and Ocker 2003). However, negative politeness seems to be more frequent in CMC than in face-to-face interaction (Carlo and Yoo 2007).

One of the topics that has attracted the most interest is gender differences in politeness. Thus, Herring (1994) reports “a tendency for women to favour positive politeness and men negative politeness,” although the most remarkable difference she finds is that flaming (i.e., posting angry or insulting messages) is “practised almost exclusively by men” (Herring 1994: 291). Similar conclusions are reached by Hall (1996) and Herring (1996; 2000), who also suggest that, while women tend to be more worried about politeness, men tend to engage in more FTAs and “to be more concerned about threats to freedom of expression than with attending to others’ social ‘face’” (Herring 2000: 3). Similarly, Guiller and Durndell (2006) found that, in educational forums, males tend to use more authoritative language and argumentation than females. However, Herring (1996) also suggests that these gender differences may disappear in mixed-group forums where members of the minority gender tend to imitate the majority gender communicative style. Likewise, Savicki et al. (1996) and Tet Mei et al. (2023) show that CMC is gradually becoming more gender-neutral in terms of politeness features, possibly because participants tend to accommodate each other’s gendered language styles (Thomson and Murachver 2001).

Assuming the existence of gender differences in studies of language use is, however, controversial. In fact, the pre-conception that women and men can be viewed as internally homogeneous groups has been progressively abandoned in the feminist literature (Cameron 1992). According to ‘the dynamic approach’ to gender (West and Zimmerman 1987; Crawford 1995) gender is not “a static, add-on characteristic of speakers, but is something that is accomplished in talk every time we speak” (Coates 2004: 7). In addition, exploring gender differences in the context of CMC research is criticised on the grounds that CMC possesses a “degree of anonymity that makes the gender of online communicators irrelevant or invisible” (Graddol and Swann 1989, as cited in Herring and Stoerger 2014: 567). In contrast, Yates (2001) argues that the gender differences found in face-to-face research are sometimes magnified in CMC, since “gender is often visible in CMC on the basis of features of a participant’s discourse style” (Guiller and Durndell 2006: 368). Additionally, Herring and Stoerger (2014: 576) remark that most instances of asynchronous CMC are not anonymous and, even when pseudonyms are used, gender can still be identified since “communicators give off cues through their interactional style and message content.”

Other issues dealt with in the CMC literature on politeness include differences in politeness in CMC versus non-CMC discourse (Brysbaert and Lahousse 2019), the relationship between politeness and discourse functions (Schallert et al. 2009), and the effects of time in the communicative and politeness practices of online learning communities, and L1-related differences in strategy choice (Fernández-Polo and Cal-Varela 2017; Cal-Varela and Fernández-Polo 2020).

1.2. Argumentative concessives and politeness

Despite the important argumentative value ascribed to concessive connectors in the literature on academic discourse (Biber et al. 1999; Couper-Kuhlen and Thompson 2000), little research has been conducted on the role played by these rhetorical relations in CMC. A few exceptions can be found. Tanskanen and Karhukorpi (2008), for instance, explore how participants in e-mail conversations use concessives to correct themselves. Their study suggests that when participants use concessives to repair claims that may cause disagreement, they are adopting the perspective of their fellow communicators and negotiating affiliation “in a dialogical manner” (Tanskanen and Karhukorpi 2008: 1587). Drawing on an interest in online forums as channels for public dialogue on current political and social issues, Swanson et al. (2015) deal with concession in the context of argument mining. Thus, they suggest that statements containing “specification, contrast, concession and contingency markers are more likely to contain good argumentative segments” (Swanson et al. 2015: 218). Most remarkably, concessives are the focus of a study conducted by Musi et al. (2018), who test the hypothesis that argumentative concessions can be used as persuasive strategies by calculating their frequency in persuasive vs. non-persuasive discourse. For this purpose, they use a CMC dataset, the ChangeMyView Subreddit platform, “where multiple users negotiate opinions on a certain issue willing to change their point of view through other users’ arguments” (Musi et al. 2018: 2). Although their results suggest that concessions do not make the arguments more convincing in this specific context, they argue that this is because their persuasive value is “context-bounded and crucially depends on the rhetorical situation” (Musi et al. 2018: 16).

Outside the CMC context, the literature on concessives has referred indirectly to their role as politeness strategies. Thus, Biber’s (1988) multidimensional approach associates concessives with other mitigating devices such as hedges or downtoners; in this model, concession is a marker of non-assertiveness, since it indicates the possibility that other options are true (Monaco 2017: 138). Furthermore, it is often mentioned that concessives are used to increase the hearer’s positive attitude towards the speaker’s opinion (Mann and Thompson 1988), since “recognizing the validity of the hearer’s standpoint before expressing disagreement can avoid FTAs acts and is perceived as reasonable by the hearer” (Couper-Kuhlen and Thompson 2000: 381). Additionally, some studies have emphasised the correlation between the use of concessive connectors and the presence of opinion, evaluation, and argumentation (Swanson et al. 2015).

The notion of ‘concession’ used here is based on Couper-Kuhlen and Thompson’s (2000: 381) definition of concessives as three-part sequences in which: 1) the first speaker makes a point (X), 2) the second speaker concedes the validity of this point (X’), and 3) the second speaker makes a potentially contrasting point (Y). This description provides the basis for Musi et al.’s definition of ‘argumentative concessives’ (ACs) as a type of concessive in which “the proposition introduced by the connective – B –, which denies the expectations brought about by a preceding proposition, expresses the speaker’s standpoint” (Musi et al. 2018: 5). According to these authors, at a semantic level, the conceding proposition (or proposition A) of ACs typically includes agreement or a positive evaluation of the statement previously presented by the other speaker, while the denial-of expectations proposition (or proposition B) tends to include (mitigated) criticism. Additionally, Musi et al. (2018) suggest that ACs can be characterised by referring to the linguistic features that tend to co-occur with them. Their list of features includes: a) hedges (defined as lexical and syntactic means of decreasing the writer’s responsibility “for the extent and the truth-value of propositions and claims, displaying hesitation, uncertainty, indirectness, and/or politeness to reduce the imposition on the reader” (Hinkel 2005: 30)); b) positive and negative sentiment words (since ACs usually contain opinion on the other posts); c) first and second personal pronouns and adjectives (since ACs “dialogically point to the stance taken by the previous speaker” (Musi et al. 2018: 10)); and d) modal verbs, which indicate that what is expressed in proposition B is not ‘unassailable.’

1.3. The current study

This project aims to provide a preliminary characterisation of concessives in the Santiago University Corpus of Discussions in Academic Contexts (SUNCODAC 2021) and to show the relevance of politeness for this characterisation. More specifically, we are interested in exploring how L1-Spanish EFL learners participating in this discussion forum use but-concessives (henceforth, butCs) for argumentation.

The decision to include only butCs was motivated by the fact that these connectives have been found to be the most frequent concessive marker in different discourse types (Grote et al. 1997; Izutsu 2008; Taboada and Gómez-González 2012; Gómez-González 2017).² Additionally, but represents 85 per cent of concessive markers in discussion forums (Musi et al. 2018) and 52 per cent of all concessive markers in the English component of SUNCODAC (Doval-Suárez and González-Álvarez 2021). Barth (2000: 418) explains that the reasons for this prevalence of but are not only connected with the fact that they are paratactic constructions which “facilitate on-line production,” but also, and most importantly, with the fact that they “provide an opportunity for face work by leaving the speaker room to manoeuvre and by attending to the recipient’s need for politeness.” Additionally, Uzelgun et al. (2015) suggest that the yes … but-construction plays a key role in the study of (dis)agreement space by presenting what is accepted as opposed to what is criticised.

Drawing on Musi et al.’s (2018) characterisation of concession, we focused on the use of butCs in combination with hedges, positive and negative words, and first and second personal pronouns and adjectives.³ Additionally, boosters were also included as a category in this characterisation. The reason for this is that, together with hedges, boosters can function as stance markers or markers of epistemic modality, since they are used by a speaker/writer “to signal different degrees of certainty concerning the validity of the information” and “to increase or decrease the illocutionary force of speech acts” (Holmes 1982: 11). Therefore, boosters and hedges are two sides of the same coin.

2. Method

2.1. Participants and data source

SUNCODAC, the corpus used in this study consists of student forum discussions gathered over a span of four years at the University of Santiago de Compostela (USC).⁴ These discussions were an integral part of an English-to-Spanish translation course designed for second-year undergraduates, primarily majoring in English at USC. The forum served as a supplementary tool alongside traditional face-to-face teaching, and students actively contributed at three distinct times during the semester: the beginning (period 1), middle (period 2), and end (period 3).

As shown in Table 1, the corpus contains a representation of English, Spanish, and Galician used as first (L1) and second (L2) languages by students of different nationalities. The subjects are L1 and L2 English speakers of several L1 backgrounds, mainly Spanish, Galician, English and Chinese, but this study concentrates on L1-Spanish participants’ productions in L2 English.

Languages	Words	Posts		Number of participants
Spanish	232,440	1,521	Gender	L1 Sp./Gal.	L1 English	L1 Chinese	L1 Other	Total
Galician	18,547	119	Female	295	17	56	30	398
English	328,537	1,724	Male	87	8	20	7	122
Total	579,524	3,364	Total	382	25	76	37	520

Table 1: A description of SUNCODAC

A detailed description of the activity can be found in Cal-Varela and Fernández-Polo (2020: 46–47). Every week, a practical session was allocated for in-class discussions on a translation topic, followed by an online discussion. To facilitate this, distinct weekly forums were created within the Moodle platform. Each forum was overseen by a student who was assigned the role of moderator. The activity unfolded through five stages:

1. Lecturers’ instructions. A single opening post by the lecturers including the source text, the moderator’s name, basic instructions, and deadlines.
2. Moderator’s first translation.
3. Peer feedback. This is the core of the discussion and consists of messages where the moderator’s classmates make comments and suggestions for improvement and discuss the suitability of different translation solutions.
4. Moderator’s improved version and summary of discussion.
5. Instructor’s assessment and appraisal of the activity.

It should be noted that most of the corpus consists of feedback messages, that is, posts belonging to stage 3. Therefore, posts from this stage are the central part of the discussion and the bulk of the corpus. Each of these feedback texts may have different sections or moves (Fernández-Polo and Cal-Varela 2018):

1. Pre-proposal: provides an overall evaluation of the translation, may touch upon potential weak points, mention other aspects like task difficulty, or include expressions of congratulations.
2. Proposal: represents the core of the message, listing problems in the translation provided and offering suggestions for improvement.
3. Post-proposal: is often quite similar to the pre-proposal (but appears less frequently).
4. Opening and Closing sections: these two sections exhibit an epistolary style. The opening section features a salutation, and the closing section includes various expressions of farewell.

The different sections in peer feedback posts are illustrated in (1). ⁵

1. OPENING
   Hi everyone!
   PRE-PROPOSAL  
   I think that your translation is very good, but I would change a couple of things.
   PROPOSAL
   For example, instead of “porque afecta a la recuperación de las heridas.” I put “ya que afecta a la recuperación de lesiones” because I think that it refers to a general term (lesiones). Then, in “Este líquido necesita ser reemplazado rápidamente para contribuir a la recuperación de las articulaciones doloridas y de los músculos” I put “Este fluído debe ser reemplazado rápidamente para eliminar los dolores en las articulaciones y músculos” because it sounds more natural, more like a colloquial language.
   POST-PROPOSAL
   For the rest my translation is the same as yours, so that's all.
   CLOSING
   Regards!

(16MPU_ The best food for footballers 2016-A)

2.2. Procedure

Since this is a small-scale study, the first step was to create a subcorpus of butCs. Therefore, using the corpus search tool, a sample of 165 butCs produced by the L1-Spanish group was extracted from the English component of SUNCODAC feedback messages This sample represents 15 per cent of the overall occurrence of this marker in the whole corpus. The butCs are uniformly distributed across sections, gender groups and periods, i.e., we selected equal numbers of butCs for each level of the different variables used as corpus design criteria: gender, post section, and post period.

The creation of the subcorpus was followed by the automatic extraction of examples containing the different lexical features under study using Wordsmith Tools 7 (Scott 2016), and by the manual disambiguation of examples. The list of lexical features was constructed by referring to previous studies. Thus, we used the lists of hedges found in Hyland (2005), the list of intensifiers used by Hinkel (2005), and, in order to select the positive and negative sentiment words, we chose the sentiment/opinion lexicon published by Hu and Liu (2004), also adopted by Musi et al. (2018).

The variables considered in the subsequent quantitative analyses were the concessive proposition (A/B), the post section, course period, and gender.⁶ The quantitative analyses used Log Likelihood to test for statistically significant differences.

2.3. Research questions

In order to describe how a specific type of concessive (i.e., butC) is used in combination with other politeness strategies in a specific CMC mode (i.e., online discussion forums), our study addresses the following five research questions:

1. How frequently do butCs co-occur with the following lexical features: boosters, hedges, positive and negative sentiment words, and first and second personal pronouns and adjectives?
2. What is the distribution of these lexical features in propositions A and B of the concessive?
3. Are there any significant differences in the frequency and distribution of these lexical features between message sections (preproposal/proposal)?⁷
4. Are there any significant differences in the frequency and distribution of these linguistic features between butCs produced at the beginning and the end of the term (i.e., period 1 and period 3)?
5. Are there any significant differences in the frequency and distribution of these linguistic features between butCs produced by male and female participants?

3. Results and discussion

3.1. Towards a characterisation of concessives in SUNCODAC

The first step in the characterisation of concessives involved checking whether butCs in SUNCODAC followed the interactional and semantic patterns described by Couper-Kuhlen and Thompson (2000: 38). Our analysis revealed that most butCs in SUNCODAC typically form part of a tripartite sequence in which: a) Student 1 posts a translation, i.e., makes a point (X); b) in another post, Student 2 concedes the validity of the other student’s point in proposition A (the conceding move) by means of partial agreement, approval or praise for the proposed translation (X’);  and c) Student 2 goes on to make a potentially contrasting point in proposition B (the denial of expectation move) by suggesting changes to the original translation (Y). Additionally, and drawing on Musi et al. (2018)’s semantic characterisation, our butCs were found to consist of a conceding move containing positive sentiment or agreement in proposition A and a denial-of-expectation move containing some sort of mitigated criticism or imposition in proposition B (cf. Figure 1).

Figure 1: Typical concessive pattern (i.e., pattern 1) in SUNCODAC

However, the analysis revealed that this pattern (henceforth, pattern 1), though prevalent in the corpus, could not account for all the instances of butCs. Thus, a corpus-based approach was adopted to detect other interactive/semantic patterns. As a result of the manual analysis of concordance lines, two additional patterns emerged, whose respective frequencies are shown in Table 2.

	Number	Percentage
Pattern 1	124	75.2
Pattern 2	19	11.5
Pattern 3	22	13.3
Total	165	100.0

Table 2: Concessive patterns in SUNCODAC butCs

Figure 2 shows that, in pattern 2, which represented 11.5 per cent of the instances of butCs, the order was occasionally reversed so that proposition A was the one including the alternative translation, while proposition B was the one containing positive evaluation:

Figure 2: Alternative concessive pattern in SUNCODAC

Finally, a miscellaneous pattern (pattern 3) was also identified to account for variations of the preceding two patterns as in (5), where proposition A includes the alternative translation and B is an evaluation of this alternative. Another example is (6) where a butC appears inside another concessive headed by although. This heterogeneous pattern 3 represents 13.3 per cent of the total tokens of butCs.

5. “Finally, it sounds better for me “largas mensulas piramidales invertidas”, but maybe it is a bit stiff.” (16DRP_Male_The gift of the gab_2016A)
6. “Although this is a good translation, I would use “intentar” instead of “tratar”, but it is just because it sounds more casual for me.” (17AGO_The river_2017-A)

The final step followed in order to describe concessives in our corpus involved exploring the potential co-occurrence of butCs with boosters, hedges, positive and negative sentiment words, and first and second personal pronouns and adjectives (henceforth, I-words, you-words and we-words). In order to determine the importance of these elements for their characterisation, two measures were used: a) the proportion of butCs including each of these features (cf. Table 3), and b) their distribution in propositions A and B (Table 4).

	Number of concessives	Percentage of concessives
I-words	145	87.9
Positive	106	64.2
You-words	93	56.4
Hedges	85	51.5
Boosters	60	36.4
Negative	19	11.5
We-words	15	9.1
Zero features	19	11.5

Table 3: Frequency of butCs containing at least one token of each of the selected linguistic features

The data in Table 3 show that only 11.5 per cent of butCs in our corpus exhibit no examples of the linguistic features under consideration. As for the butCs containing at least one token of each of these features, their frequencies are presented in decreasing order: 87.9 per cent of the butCs contain at least one I-word, which highlights the importance of first-person voice. A similar incidence of ‘egocentric deictic reference’ has been detected in other forms of CMC, such as corporate blogs (Puschman 2010: 181), where participants are likely to feature prominently in their own discourse. In contrast, butCs with we-words are placed much lower in the rank (9.09%), but are also noteworthy, since they are sometimes used as a positive politeness strategy with the purpose of including the writer and the reader in the activity, thus reinforcing the sense of community of learning (7), an effect that can be also achieved by using a combination of I- and you-words, as in (8). In other cases, we-words are simply used as an instance of generalisation (9) or as negative politeness strategies to impersonalise an imposition, as in (10):

7. May I start saying that is a great translation, but I completely agree with the suggestions our mates made, like 17DVM ‘s, to make the text more natural (17ARB_The river_2017A)
8. I like you, wrote that the little door behind the curtain was 40 cm tall, but I read some of our classmates answers and I have to agree with the ones that put instead something in the lines of “de dos palmos de altura” or something of the sort. (16ASE_Alice in Wonderland_2016-A)
9. From my point we can consider your translation as more “technique” in the sense that you are using specific lexicon, but here we are writing in a newspaper and the most important thing is to arrive to the greatest possible number of people. (17AHF_Smart jacket_2017)
10. I think that translating “who” by “los cuales” does not sounds too formal, it’s a relative more complex than “que” but in this context we can use “que”, “los cuales” or “quienes” because they tree have the same meaning in here. (16ACC_ English on the march_2016-B)

The percentage of concessives including instances of second person reference is smaller but still important (56.4%) in a context characterised by appeals to other users. Furthermore, if I-words, you-words and we-words are considered together, their high prevalence may point to the dialogical character of discussion forums, which, like other types of CMC are said to “bristle with first and second person pronouns” (Jonsson 2015: 215).

Leaving out pronouns, the most characteristic feature of ACs seems to be the presence of at least one positive word or a hedge in more than half the examples. On the one hand, positive words, which appear in 64.2 per cent of the concessives, are used for subjective evaluation and are an indication of the presence of positive politeness strategies such as praise, appreciation, or gratitude (Schallert et al. 2009: 718). On the other hand, the appearance of hedges in 51.5 per cent of the examples may suggest a cautious, non-assertive kind of discourse. Hedges in our corpus are typically used as negative politeness strategies, i.e., in order to minimise the imposition represented by the suggested changes to the original translation (11).

11. Hello! I agree with you, 16VVE, but I would change the translation. (16AFF_Emergency_2016-A)

Finally, the high incidence of hedges and positive words contrasts with the relatively low percentage of butCs containing boosters and negative words, both of which fall considerably below the halfway point (with percentages of 36.4% and 11.5% respectively). A tentative explanation may be that boosters often help participants to construct a more authoritative or confident kind of discourse, which conflicts with the attenuation effect of hedges. Furthermore, the predominant function of the examined concessive clauses seems to be mitigation rather than the overt expression of criticism by means of negative sentiment. In fact, as could be seen in (5), repeated as (12) for convenience, and (13) below, many of the negative words in the corpus are not used to criticise the other participant’s translation, but to evaluate the speaker’s own proposal, in an attempt to diminish the FTA of imposing a change:

12. Finally, it sounds better for me “largas mensulas piramidales invertidas”, but maybe it is a bit stiff. (16DRP_The gift of the gab_2016-A)
13. It sounds too much formal and non-natural for me the first time I read it. So sorry. But doesn’t matter, it’s just my stupid opinion. (16DRP_Emergency_2016-B)

The final step in the characterisation of ACs involved exploring the overall frequency of each of these features in the but-corpus and their distribution in the two concessive propositions (A and B). For this purpose, we calculated their raw and relative frequencies, as shown in Table 4.

	Overall		Proposition A		Proposition B		Log Likelihood (LL)
	Raw	Fpttw	Raw	Fpttw	Raw	Fpttw
I-words	273	453.5	127	466.2	146	443.0	+0.18
You-words	129	214.3	95	348.8	34	103.2	+42.83**
Positive	127	211.0	94	345.1	33	121.2	+43.32**
Hedges	101	167.8	24	88.1	77	282.7	-20.07**
Boosters	70	116.3	48	68.6	22	31.4	+15.48**
Negative	20	33.2	9	3.0	11	40.4	-0.00
We-words	17	28.2	7	25.7	10	30.3	-0.11

Table 4: Distribution of hedges, negative and positive sentiment words in the two concessive propositions

The results of the Log Likelihood test indicate the existence of statistically significant differences (**) between the two concessive propositions regarding the frequency of you-words, positive sentiment words, hedges and boosters. On the one hand, you-words and positive words are significantly more frequent in proposition A (LL=+43.32; p<.05 and LL=+42.83; p<.05, respectively), since this is the one usually containing some sort of (boosted) praise of the other participant’s translation. An illustration of the occurrence of you-words and positive sentiment in proposition A can be found in (14) and (15) below:⁸

14. Hi everybody! I think that your translation, 16MSF, is excellent but I have some differences (16ASP_Alice in Wonderland_2016-A).
15. First of all, congratulations to you 16NBA, you have done a wonderful job translating this text, but I would like to point out some things that I translated differently. (16AEG_ The best food for footballers_2016-A)

On the other hand, our findings indicate that hedges are overused in proposition B (LL=-20.07; p<.05), which confirms the tendency observed by Musi et al. (2018), who refer to the frequent presence of modal verbs, a specific type of hedge, in proposition B of argumentative concessions. This is coherent with the fact that proposition B is the one presenting the improvements to the original translation made by the other student. In line with Hyland (1996), the presence of hedges might serve to make this proposal easier to accept by softening its tone. This can be seen in examples (16) and (17):

16. Hi 16VPL ! You have done a great translation, but I think some things could be changed (16AFF_Shrinking families_2016-A).
17. I think you have done a great job with your translation but I have some suggestions that perhaps, they may help you (or not)” (17AAR_Cellscope Oto_2017-B)

As for negative words, the results are inconclusive, since no statistically significant differences were found between the two propositions. In other words, although negative sentiment is slightly more frequent in proposition B (18), it seems that its presence in proposition A is not necessarily connected with the evaluation of the translation under discussion. For instance, the word complicated in (19) is not used to qualify the other student’s translation, but to highlight that they did a good job despite the difficulty of the task:

18. I understand that you are trying to keep the original tone of the text, but as the others said, it sounds weird here. (16DRP_The best food for footballers_2016-A)
19. Firstly, I wanted to say I think you did a good job translating this extract of the text since I found it a bit complicated, but there are some words and expressions I’d change. (17AGG_The river_2017-A)

We have, therefore, established a tentative characterisation of concessives in our corpus in terms of some of the linguistic features that co-occur with them. The following sections will explore if these findings can be further qualified by considering (contextual) factors such as course period, message section, and participant’s gender.

3.2. Frequency and distribution of lexical features in butC in relation to different contextual factors

3.2.1. Message sections

We have already described the posts as consisting of different sections or moves. According to the corpus compilers (Fernández-Polo and Cal-Varela 2018), forum posts in SUNCODAC can be categorised as a genre, understood as comprising standard sequences of moves, or text segments that play identifiable roles within the overall structure. The authors found that the structural components observed in SUNCODAC exhibit move-like properties, characterised by distinctive language and specific text positions (Fernández-Polo and Cal-Varela 2018: 192).

To determine whether the two sections containing butCs in our subcorpus also have distinctive language characteristics, we conducted a comparison of the frequency of the features considered for our study between the preproposal and proposal sections. The results are shown in Table 5.

	Proposal		Proposal		Log Likelihood (LL)
	Raw	Fpttw	Raw	Fpttw
Hedges	60	350.7	41	156.8	+16.16**
Boosters	14	81.8	56	214.1	-12.29 **
Negative	2	11.7	18	68.8	-8.83**
Positive	60	350.7	71	271.5	+2.11
I-words	149	935.1	126	443.6	+23.96**
You-words	92	420.8	37	221.8	+53.31**
We-words	8	46.8	9	34.4	+0.39

Table 5: Frequency and distribution of lexical features of butCs appearing in different message sections

As shown in Table 5, the Log Likelihood test reveals a significantly higher frequency of hedges, I- and you-words in preproposals, but a significantly lower frequency of boosters and negative sentiment words. No significant differences were observed in the frequency of use of positive words and we-words.

The fact that hedges in butCs are significantly more frequent in preproposals than in proposals can be explained by the fact that preproposals in SUNCODAC are often used to announce criticism, and hedges often co-occur with critical comments, precisely because of their ability to keep the distance between what is being said and the actual writer’s opinion. Thus, any conflicts that could arise from explicit claims to an absolute truth are avoided. Similarly, Cal-Varela and Fernández-Polo (2020) identified hedges as part of the mitigating strategies of preproposals in the Spanish subcomponent of SUNCODAC.

While hedges are more frequent in preproposals, boosters and negative words are significantly more frequent in proposals, whereas positive words and we-words appear with similar frequencies in both types of sections. A tentative interpretation of these results is that in proposals the focus is on criticism. Thus, boosters are used for two apparently contradictory purposes: to boost criticism, qualifying the writer’s commitment to the truth of the proposition; and to project added politeness, sincerity and truthfulness. As for negative words, it has already been observed (cf. Section 3.1.) that they are sometimes used to qualify the speaker’s own proposal, serving as negative politeness strategies that reduce the force of the imposition caused by suggesting an alternative translation.

On the other hand, positive words are slightly overused in proposals, which is in line with Tan et al. (2016), who state that persuasive opening arguments (which could be the equivalent of SUNCODAC preproposals) use fewer positive words, suggesting more complex patterns of positive emotion in longer arguments appearing later in the message (i.e., proposals). However, this finding should be taken with care since this overuse is not statistically significant.

3.2.2. Course period

It has been observed that the presence of politeness devices, which serve to mitigate critical comments and contribute to a more congenial learning environment, might be anticipated to evolve over the duration of the course, particularly if the group develops into a genuine community of inquiry (Fernández-Polo and Cal-Varela 2017: 260). To investigate whether this evolution occurs in the forum discussions being analysed, we conducted a comparison of the frequency of different features between Period 1 and Period 3. The results can be seen in Table 6.

	Period 1		Period 3		Log Likelihood (LL)
	Raw	Fpttw	Raw	Fpttw
Hedges	51	324.8	30	207.0	+ 3.95 **
Boosters	19	121.0	28	193.2	- 2.53
Negative	1	6.4	1	6.9	- 0.00
Positive	46	293.0	19	131.1	+ 9.50 **
I-words	109	694.3	79	545.2	+ 2.70
You-words	38	242.0	49	338.2	- 2.42
We-words	6	38.2	7	48.3	- 0.18

Table 6: Frequency and distribution of lexical features of butCs in first and last course periods

Table 6 shows a significant decrease in the use of hedges and positive sentiment words over the time span of the course, which can indicate that as participants get to know each other, they feel less need to mitigate the force of the criticism. On the contrary, although not significant, there is a decrease in the use of I-words, and an increase in the frequency of you- and we-words, which could point to an evolution from a “mostly monologic, informational and author-centred” kind of post to “a progressively longer post with […] a heightened awareness of the dialogic and multi-party nature of the exchanges” (Fernández Polo and Cal-Varela 2017: 256).

An evolution over the time span of the course, but pointing in the opposite direction, was observed in previous studies on the use of mitigation strategies in CMC, where intensity by repetition of the same strategy and attenuation effort (measured by the combination of different attenuating strategies) seem to increase with time, “suggesting that students become increasingly aware of the need to step up interpersonal work” (Cal-Varela and Fernández-Polo 2020: 50).

3.2.3. Gender

Contradictory findings have been obtained in previous studies of the influence of the gender factor on the use of mitigation strategies in discussion forums (cf. Section 1.1). Table 7 shows the distribution of features in relation to gender in the present study.

	Male participants		Female participants		Log Likelihood (LL)
	Raw	Fpttw	Raw	Fpttw
Hedges	30	154.1	71	298.4	- 9.93**
Intensifiers	16	80.1	54	227.0	-15.67**
Negative	18	90.1	2	8.4	+17.68**
Positive	76	390.3	55	231.2	+ 8.9 **
I-words	97	498.2	179	752.4	- 12.49**
You-words	32	164.4	98	411.9	-24.56**
We-words	3	15.4	14	58.9	-5.93**

Table 7: Frequency and distribution of hedges and positive words in butCs produced by male and female participants

The Log Likelihood tests show statistically significant differences between male and female participants. Hedges, boosters, and pronouns are significantly over-represented in females’ posts, whereas positive and negative sentiment words are significantly over-represented in posts produced by male participants. The significantly higher frequency of hedges in women’s posts seems to confirm assumptions that females tend to use more attenuated speech forms (cf. Guiller and Durndell 2006; Hall 1996; Herring 1994; 1996; 2000), since these features can be used both to attenuate criticism (positive politeness) or to attenuate imposition (negative politeness). Additionally, the higher frequency of personal pronouns in females’ posts could be interpreted as an indication of a more dialogical kind of discourse. Finally, considering that the connotations of boosters can vary based on the words they modify, a more in-depth qualitative investigation is necessary for a nuanced interpretation of the findings.

4. Conclusions

As shown in the previous sections, addressing the description of concessives by referring to their role as politeness strategies is especially relevant for the study of CMC contexts which include assessment and evaluation of peers’ writing. The characterisation of butCs carried out in this study was approached in several steps. First, we identified the interactive patterns that are typically followed by butCs and concluded that they usually stick to the structures previously mentioned (Couper-Kuhlen and Thompson 2000; Musi et al. 2018). This means that proposition A is semantically characterised by the presence of praise or agreement with the other student’s translation, and proposition B is semantically characterised by the presence of mitigated imposition. However, other patterns emerged from the study, which call for a more detailed analysis including more examples and other concessive connectors.

Then, we explored the relative importance of the co-occurrence of butCs with first and second personal pronouns and adjectives, hedges, boosters, and positive and negative words by considering the frequency of this co-occurrence. The fact that only a small percentage of butCs contains no instances of these features seems to indicate that their presence is highly relevant in this characterisation. Furthermore, the high overall incidence of butCs containing I- and you-words points to a type of discourse in which the high prevalence of first-person voice combines with the importance of the appeal to other users, as happens in texts of a dialogical nature. Also, the abundance of butCs with positive words and hedges suggests that participants in these discussions are focused on “phrasing things in such a way as to take into consideration the feelings of others” (Morand and Ocker 2003: 2). This concern for politeness becomes particularly important in a context where the interactions typically involve assessing each other’s production.

Additionally, our findings reveal that the typical distribution of these lexical features in SUNCODAC butCs is clearly determined by the proposition, and that this distribution directly mirrors the semantic and interactional function of each proposition. Thus, boosters, you-words and positive sentiment words feature prominently in proposition A, while proposition B is clearly marked by the presence of hedges. In contrast, no statistical differences were found in the case of negative words and boosters, whose low frequency may be connected with the fact that SUNCODAC participants tend to avoid overt criticism of the other participants’ translations (i.e., they try to minimise threats to positive face), and also avoid presenting their alternative translations in a way that can be perceived as a threat to their classmates’ negative face (hence the occasional use of negative words to qualify their own suggestions for improvement).

We contend that the emerging characterisation is relevant for Brown and Levinson’s model of politeness for three reasons: a) proposition B typically contains a FTA, i.e., an imposition realised as a suggestion for improvement of another student’s translation, b) this imposition is typically mitigated by means of hedging, an example of the workings of negative politeness, and c) the FTA in proposition B is typically preceded in proposition A by some sort of positive politeness realised as positive evaluation or agreement with the other student. Furthermore, this characterisation may afford a new insight into the use of politeness strategies not only in asynchronous online discussion forums but also in other CMC modes as well.

In order to address research questions 3, 4 and 5, we investigated if the previous characterisation could be further qualified by considering two task-related factors (message-section and course period) and one participant-related factor (gender). The attested variations in the frequency of occurrence of the different lexical features in the two message sections indicate that these features can be used to characterise preproposals and proposals as distinct moves. Again, the fact that hedges and I- and you-words are significantly overused in preproposals, while boosters and negative words are significantly overused in proposals, mirrors the functions each section has in the post, which is in line with results obtained in previous studies (Fernández-Polo and Cal-Varela 2018). As for the effects of time, our results clearly point to a significant decrease in the use of hedges and positive words over time, which might imply that as the term progresses, participants are less concerned about politeness issues. We also traced an increase in the use of you- and we-words which could be a symptom of a gradual evolution towards a more dialogic and multi-party type of discourse.

The existence of gender-based differences in the use of politeness strategies is by no means uncontroversial, and our findings regarding this issue are rather inconclusive. On the one hand, the existence of significant differences between male and female participants in the frequency of the different lexical features might suggest that they have different styles. Thus, our results show that females significantly favour hedges, boosters and pronouns, while men favour the expression of both positive and negative sentiment. In terms of politeness strategies, women tend to mitigate more while men seem to praise more, but also to impose or criticise more often. On the other hand, the overrepresentation of hedges in posts written by female participants (which suggests that they use a more attenuated style) seems to conflict with the fact that they also overuse boosters (an indication of assertiveness), which points to the need to adopt a different perspective in the study of gender and politeness. However, given the small size of the sample used in the study, these results need to be taken with care, and should be tested on a larger and more representative number of examples, which calls for a larger-scale study with a more balanced representation of male and female participants.

While awaiting the bigger picture, we have brought forward a preliminary characterisation of butCs in SUNCODAC, with some features yielding a neater description than others and some variables clearly being more significant than others. Future analysis should reveal the extent to which this characterisation can be extended to other types of concessives. In addition, further research will necessarily involve a refinement of the lists of lexical features which are relevant for the characterisation of concession. All in all, we have described how concession and other politeness strategies work together towards “creating a comfort zone in which to exchange ideas as well as motivating students’ participation” (Schallert et al. 2009: 715) in the discussion and, hence, in the learning process. We believe that our study has contributed to a better understanding of the role of this rhetorical relation in discussion forums, but its role in other types of CMC still needs to be investigated.

5. References

Barth, Dagmar. 2000. That’s true, although not really, but still: Expressing concession in spoken English. In Elizabeth Couper-Kuhlen and Bernd Kortmann eds. Cause-Condition-Concession-Contrast: Cognitive and Discourse Perspective. Berlin: Mouton de Gruyter, 411–437.

Biber, Douglas. 1988. Variation across Speech and Writing. Cambridge: Cambridge University Press.

Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad and Edward Finegan. 1999. Longman Grammar of Spoken and Written English. London: Longman.

Brown, Penelope and Steven C. Levinson. 1978. Universals in language usage: Politeness phenomena. In Esther N. Goody ed. Questions and Politeness. Cambridge: Cambridge University Press, 56–289.

Brown, Penelope and Stephen C. Levinson. 1987. Politeness: Some Universals in Language Usage. Cambridge: Cambridge University Press.

Brysbaert, Jorina and Karen Lahousse. 2019. Computer-mediated versus non-computer-mediated corpora of informal French: Differences in politeness and intensification in the expression of contrast by au contraire. In Julien Longhi and Claudia Marinica eds. Proceedings of the 7^th Conference on CMC and Social Media Corpora for the Humanities. Cergy-Pontoise: The Institute Of Digital Humanities of Cergy-Pontoise University, 48–52.

Cal-Varela, Mario and Francisco Javier Fernández-Polo. 2019. Preparing the ground for critical feedback in online discussions: A look at mitigation strategies. In Julien Longhi and Claudia Marinica eds. CMC Corpora through the Prism of Digital Humanities. Paris: L’Harmattan, 15–34.

Cal-Varela, Mario and Francisco Javier Fernández-Polo. 2020. SUNCODAC: A corpus of online forums in higher education. Nexus-AEDEAN 2: 44–52.

Cameron, Deborah. 1992. ‘Not gender difference but the difference gender makes’: Explanation in research on sex and language. International Journal of the Sociology of Language 1992/92: 13-26.

Carlo, Jessica Luo and Youngjin Yoo. 2007. “How may I help you?” Politeness in computer-mediated and face-to-face library reference transactions. Information and Organization 17/4: 193–231.

Coates, Jennifer. 2004. Women, Men and Language: A Sociolinguistic Account of Gender Differences in Language. New York: Routledge.

Couper-Kuhlen, Elizabeth and Sandra Thompson. 2000. Concessive patterns in conversation. In Elizabeth Couper-Kuhlen and Bernd Kortmann eds. Cause, Concession, Contrast: Cognitive and Discourse Perspectives. Berlin: Mouton de Gruyter, 381–410.

Crawford, Mary. 1995. Talking Difference: On Gender and Language. London: Sage.

Deris, Farhana Diana, Rachel Tan Hooi Koon and Abdul Rahim Salam. 2015. Virtual Communities in an Online English language learning forum. International Education Studies 8/12: 79–87.

Doval-Suárez, Susana M. and Elsa González-Álvarez. 2021. A Comparison of NS and NNS use of Concessive Markers in Computer- and Non-computer Mediated Discourse. Paper presented at the 9^th International Contrastive Linguistics Conference, University of Genoa (Italy), 20–21 May 2021.

Fernández-Polo, Francisco Javier and Mario Cal-Varela. 2017. A description of asynchronous online discussions in higher education. In Chelo Vargas-Sierra ed. Professional and Academic Discourse: An Interdisciplinary Perspective. Epic Series in Language and Linguistics 2. Alicante: AESLA, 256–264.

Fernández-Polo, Francisco Javier and Mario Cal-Varela. 2018. A structural analysis of student online forum discussions. In Francisco Javier Díaz Pérez and María Águeda Moreno Moreno eds. Looking at the Crossroads: Training, Accreditation and Context of Use. Jaén: University of Jaén, 189–200.

Goffman, Erving. 1967. Interaction Ritual: Essays on Face-to-Face Interaction. Chicago: Aldine.

Gómez-González, María de los Ángeles. 2017. Concession in evaluative discourse: The semantics, pragmatics and discourse effects of ‘but’ and ‘although’. In Ruth Breeze and Inés Olza eds. Evaluation in Media Discourse: Rhetoric and Intercultural Pragmatics. Bern and New York: Peter Lang, 47–80.

Graddol, David and Joan Swan. 1989. Gender voices. Cambridge: Basil Blackwell.

Grote, Brigitte, Nils Lenke and Manfed Stede. 1997. Ma(r)king concessions in English and German. Discourse Processes 24/1: 87–117.

Guiller, Jane and Alan Durndell. 2006. ‘I totally agree with you’: Gender interactions in educational online discussion groups. Journal of Computer Assisted Learning 22: 368–381.

Hall, Kira. 1996. Cyberfeminism. In Susan Herring ed. Computer-Mediated Communication: Linguistic, Social and Cross-Cultural Perspectives. Amsterdam: John Benjamins, 147–170.

Harrison, Sandra. 2000. Maintaining the virtual community: Use of politeness strategies in an email discussion group. In Lyn Pemberton and Simon Shurville eds. Words on the Web. Exeter: Intellect Ltd., 69–79.

Herring, Susan C. 1994. Politeness in computer culture: Why women thank and men flame. In Mary Bucholtz ed. Cultural Performances: Proceedings of the Third Berkeley Women and Language Conference. Berkely: Berkeley Women and Language Group, 278–294.

Herring, Susan C. 1996. Posting in a different voice: Gender and ethics in computer-mediated communication. In Charles Ess ed. Philosophical Perspectives on Computer-Mediated Communication. Albany: SUNY Press, 115–145.

Herring, Susan C. 2000. Gender differences in CMC: Findings and implications. Computer Professionals for Social Responsibility Journal 18/1: 1–9.

Herring, Susan C. 2023. Grammar and electronic communication. In Carol A. Chapelle ed. The Encyclopedia of Applied Linguistics. New York: Wiley-Blackwell, 1–9.

Herring, Susan C. and Sharon Stoerger. 2014. Gender and (A)nonimity in computer- mediated communication. In Susan Ehrlich, Miriam Meyerhoff and Janet Holmes eds. The Handbook of Language, Gender, and Sexuality. Malden: Wiley-Blackwell, 567–586.

Hinkel, Eli. 2005. Hedging, inflating, and persuading. Applied Language Learning 15/1–2: 29–53.

Holmes, Janet. 1982. Expressing doubt and certainty in English. RELC Journal 13/2: 9–28.

Hu, Minqing and Bing Liu. 2004. Mining and summarizing customer reviews. In Won Kim, Ron Kohavi, Johannes Gehrke and William DuMouchel eds. Proceedings of the Tenth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Washington: ACM, 168–177.

Hyland, Ken. 1996. Talking to the academy: Forms of hedging in science research articles. Written Communication 13/2: 251–281.

Hyland, Ken. 2005. Metadiscourse. London: Continuum.

Izutsu, Mitsuku Narita. 2008 Contrast, concessive, and corrective: Toward a comprehensive study of opposition relations. Journal of Pragmatics 40/4: 646–675.

Jonsson, Ewa. 2015. Conversational writing: A Multidimensional Study of Synchronous and Supersynchronous Computer-Mediated Communication. Frankfurt am Main: Peter Lang.

Jordan, Michelle E., An-Chih Janne Cheng, Diane Schallert, Kwangok Song, SoonAh Lee and Yangjoo Park. 2014. “I guess my question is”: What is the co-occurrence of uncertainty and learning in computer-mediated discourse? International Journal of Computer-Supported Collaborative Learning 9: 451–475.

Li, Mimi. 2012. Politeness strategies in wiki-mediated communication of EFL collaborative writing tasks. The International Association for Language Learning Technology Journal 42/2: 1–26.

Locher; Miriam. 2004. Power and Politeness in Action: Disagreements in Oral Communication. New York: Mouton de Gruyter.

Locher, Miriam A. and Richard J. Watts. 2005. Politeness theory and relational work. Journal of Politeness Research 1: 9–33.

Locher, Miriam A. and Richard J. Watts. 2008. Relational work and impoliteness: Negotiating norms of linguistic behaviour. In Derek Bousfield and Miriam A. Locher eds. Impoliteness in Language: Studies on its Interplay with Power in Theory and Practice. Berlin: Mouton De Gruyter, 77–99.

Mann, William C. and Sandra A. Thompson. 1988. Rhetorical Structure Theory: Toward a functional theory of text organization. Text – Interdisciplinary Journal for the Study of Discourse 8/3: 243–281.

Monaco, Leida Maria. 2017. A Multidimensional Analysis of Late Modern English Scientific Texts from the Coruña Corpus. A Coruña: Universidade da Coruña dissertation.

Morand, David A. and Rosalie J. Ocker. 2003. Politeness theory and computer-mediated communication: A sociolinguistic approach to analysing relational messages. In Proceedings of the 36^th Hawaii International Conference on System Sciences. Hawaii: IEEE. https://doi.org/10.1109/HICSS.2003.1173660

Musi, Elena, Debanjan Ghosh and Smaranda Muresan. 2018. ChangeMyView through concessions: Do concessions increase persuasion? Discourse and Dialogue 9/1: 1–21.

Park, Jung-ran. 2008. Linguistic politeness and face-work in computer mediated communication: An application of the theoretical framework. Journal of the American Society for Information Science and Technology 59/14: 2199–2209.

Puschmann, Cornelius. 2010. Thank you for thinking we could: Use and function of interpersonal pronouns in corporate web logs. In Heidrun Dorgeloh and Anja Wanner eds. Syntactic Variation and Genre. Berlin: Mouton De Gruyter, 167–191.

Pyo, Jihoon and Chung Hyun Lee. 2019. The effects of mitigation strategies instruction in peer response to L2 writing through computer-mediated communication at university level. Multimedia-Assisted Language Learning 22/4: 103–133.

Savicki, Victor, Dawn Lingenfelter and Merle Kelley. 1996. Gender language style and group composition in Internet discussion groups. Journal of Computer-Mediated Communication 2/3, JCMC232. https://doi.org/10.1111/j.1083-6101.1996.tb00191.x

Schallert, Diane L., Yueh-hui Vanessa Chiang, Yangjoo Park, Michelle E. Jordan, Haekyung Lee, An-Chih Janne Cheng and Kwangok Song. 2009. Being polite while fulfilling different discourse functions in online classroom discussions. Computers and Education 53/3: 713–725.

Scott, Mike. 2016. WordSmith Tools Version 7. Stroud: Lexical Analysis Software.

SUNCODAC. 2021. Santiago University Corpus of Discussions in Academic Contexts. Santiago de Compostela: University of Santiago de Compostela. http://www.suncodac.com

Swanson, Reid, Brian Ecker and Marilyn A. Walker. 2015. Argument mining: Extracting arguments from online dialogue. In Svetlana Stoyancheve, Safiq Joty, David Schlangen, Ondrej Dusek, Casey Kennington and Malihe Alikhani eds. Proceedings of the 24^th Meeting of the Special Interest Group on Discourse and Dialogue. Prague: SIGDIAL, 217–226.

Taboada, Maite and María de los Ángeles Gómez-González. 2012. Discourse markers and coherence relations: Comparison across markers, languages and modalities. Linguistics and the Human Sciences 6/1-3: 17–41.

Tan, Chenhao, Vlad Niculae, Cristian Danescu-Niculescu-Mizil and Lillian Lee. 2016. Winning arguments: Interaction dynamics and persuasion strategies in good-faith online discussions. In Jaqueline Bordeau, Jim A. Hendler, Roger Nkambou, Ian Horrocks and Ben Y. Zhao eds. Proceedings of the 25^th International Conference on World Wide Web. Quebec: International World Wide Web Conferences Steering Committee, 613–624.

Tanskanen, Sanna-Kaisa and Johanna Karhukorpi. 2008. Concessive repair and negotiation of affiliation in e-mail discourse. Journal of Pragmatics 40/9: 1587–1600.

Tet Mei, Kirstie Fung, Su-Hie Ting and Kee-Man Chuah. 2023. Detecting female and male language features in Facebook comments by Malaysian millennial users. Journal of the Southeast Asian Linguistics Society 16/1: 37–55.

Thomson, Rob and Tamar Murachver. 2001. Predicting gender from electronic discourse. British Journal of Social Psychology 40/2: 193–208.

Uzelgun, Mehmet Ali, Dima Mohammed, Marcin Lewinski and Paula Castro. 2015. Managing disagreement through ‘yes, but’ constructions: An argumentative analysis. Discourse Studies 17/4: 467–484.

Van Nguyen, Long. 2010. Computer mediated collaborative learning within a communicative language teaching approach: A sociocultural perspective. The Asian EFL Journal Quarterly 12/1: 202–233.

Vinagre, Margarita. 2008. Politeness strategies in collaborative e-mail exchanges. Computers & Education 50/3: 1022–1036.

Watts, Richard. J. 1992. Linguistic politeness and politic verbal behaviour: Reconsidering claims for universality. In Richard J. Watts, Sachiko Ide and Konrad Ehlich eds. Politeness in Language: Studies in its History, Theory and Practice. Berlin: Mouton de Gruyter, 43–69.

Watts, Richard. J. 2003. Politeness. Cambridge: Cambridge University Press.

West, Candace and Don H. Zimmerman.1987. Doing gender. Gender and Society 1: 125–51.

Yates, Simon. J. 2001. Gender, language and CMC for education. Learning and Instruction 11/1: 21–34.

Notes

¹ We sincerely thank the editors and the two reviewers for taking the time to review the manuscript and providing constructive feedback which improved the original. For generous financial support, we are grateful to the following institutions: The Spanish Ministry of Science and Innovation (grant PID2021-122267NB-I00), the European Regional Development Fund (grant PID2021-122267NB-I00), and the Regional Government of Galicia (Consellería de Educación, Cultura e Universidade, grant ED431B 2021/02). [Back]

² The concessive value of but has been generally ignored in the literature. For a detailed description of the concessive, contrastive, and corrective meanings of but, see Izutsu (2008). [Back]

³ These authors also include modals as a separate category, but our study focused only on modals working as hedges. [Back]

⁴ http://www.suncodac.com/ [Back]

⁵ All examples included in the article are corpus examples which have not been altered. This means they may include spelling errors and typos, among other types of mistakes. [Back]

⁶ Although we are aware of the problematic status the category ‘gender’ (cf. Section 1), we will stick to the two-way (‘masculine’ vs. ‘feminine’) classification of the participants’ gender made by the SUNCODAC compilers. [Back]

⁷ Post-proposals are not considered here because no examples of butCs were found in this section. [Back]

⁸ In all the examples that follow, the proposition under discussion appears in bold type and the co-occurring linguistic feature is underlined. [Back]