A closer look at explanations in refusals: themes, attributions, and contexts

2.1 Explanations and refusals

Explanations usually serve to clarify an issue or situation by providing reasons or causes, with the perlocutionary effect of “enlightening someone or get[ting] him to understand” (Achinstein 1977: 1). They are commonly used to explain “the actions, beliefs and motivations of people through the identification of social causes and conditions” (Pleasants 2021: 340). By doing so, speakers can signal their prior actions as “permissible, not morally wrong”, and thus are “justified in performing it” (Pleasants 2021: 341). Heritage (1984: 271) also argued that explanations have the function of demonstrating the speakers’ “‘no fault’ quality”, thus mitigating threats to the face of either party or their relationship. Hence, a typical environment where explanations are frequently used is the dispreferred context, such as rejecting an offer or invitation. Specifically, overt explanations can perform affiliative work in refusals by “account[ing] for the trouble” (Robinson 2016: 37), framing the act as justified or even appropriate within the shared moral order. In doing so, explanations “mitigat[e] and disavo[w] negative inferences generated by problematic or dispreferred actions” (Chalfoun 2025: 64). Moreover, explanations are commonly regarded as an indirect form of refusals (Beebe et al. 1990), which, under traditional politeness theories (e.g., Brown and Levinson 1987), are assumed to be inherently polite due to their indirectness.

Numerous studies have demonstrated the prevalence of explanations in refusals across languages. For example, in a comparative study examining Chinese refusals across two contexts (formal and informal), Ren (2015) observed that speakers from the Chinese mainland and Taiwan displayed a similar preference for employing the explanation strategy in both settings, with more than a third of their overall strategies being explanations. Similarly, in Arabic and English, Nelson et al. (2002) found that reasons were the predominant refusal strategy in both languages, with Arabic speakers employing explanations more frequently than American English speakers, particularly when rejecting someone with lower social status. In Persian, Shishavan and Sharifian (2013) also demonstrated that Persian speakers heavily used explanations in their refusals, usually taking up more than 30 % in their overall indirect strategies. However, a noticeable limitation of the above studies, as well as much of the other refusal research, is that they treat explanations as a singular category, without considering the complexity and variation inherent in this strategy.

In current pragmatics literature, more detailed insights into explanations in refusals are sometimes found in a handful of cross-linguistic studies focusing on two areas: the specificity of explanations and the topics of explanations. For example, in a study comparing American English and Mandarin Chinese refusals, Liao and Bresnahan (1996: 711) noticed that Chinese participants often offered more specific reasons than their American counterparts when addressing a high-status person. Similarly, Chang (2011) found that Chinese speakers’ preference for specific reasons also exists in their EFL performance, and her participants displayed significantly more specific reasons than Americans in refusals. However, for Iranian EFL learners, Allami and Naeimi (2011) observed that their refusal explanations are less specific than American English speakers. Similarly, Beebe et al. (1990) also observed that the Japanese English speakers’ explanations are less specific compared with their American English counterparts.

In addition to specificity, previous studies also documented the variations of explanation topics across languages. For instance, when rejecting requests and invitations from a high-status person, Kwon (2004) discovered that Americans often refer to plans with partners and children, while Koreans usually mention plans with parents or the sickness of parents. Similar to Kwon (2004), Morkus (2014) also identified several topic differences between American English speakers and Arabic speakers in giving refusal explanations, with American speakers favoring personal reasons (e.g., work, study, friends) and Arabic speakers preferring family-oriented reasons and “white lie” reasons.

The above studies have shown the complexity of explanations and demonstrated that there exists great variation in Beebe et al.’s (1990: 73) category “excuse, reason, explanations”.^[2] Individuals may use explanations with different degrees of specificity and various topics in their refusals. However, most of them stop at describing the nuances of explanations at the content level. Moreover, the cross-linguistic design in much previous research often leads to the differences being attributed to cultural influence, and fails to consider other possible factors.

To move beyond surface-level content descriptions, a few studies have hinted at the relevance of attributional features in refusal explanations, even if not framed explicitly within attribution theory. For instance, in Liao and Bresnahan’s (1996) study on Chinese and English refusals, although the issue of refusal attribution was not explored explicitly, the authors briefly mentioned that Chinese participants were more likely to provide extrinsic reasons compared to their American counterparts. Similarly, Siebold and Busch’s (2015) comparative study of German and Spanish refusals categorized explanations such as lack of time as “external factors not controlled by the speaker” (Siebold and Busch 2015: 60), implicitly touching on the attributional dimensions of locus and controllability. However, neither study examined attributional dimensions in a systematic, theoretically grounded way. To this end, we turn to social psychology, drawing particularly on attribution theory to deepen our understanding of how and why refusers construct explanations. The next section introduces the theory and discusses its link with pragmatics research.

2.2 Attribution theory and pragmatics

Developed by Fritz Heider in the 1950s and further expanded by subsequent scholars like Bernard Weiner and colleagues, attribution theory “provides the framework necessary to understand how individuals explain why events in their environment happened” (Martinko and Mackey 2019: 523). According to the theory, individuals’ explanations usually encompass two key properties,^[3] including the locus of causality (internal locus: the individuals themselves vs. external locus: outside of the individuals) and controllability (controllable vs. uncontrollable) (Weiner 2010; Weiner et al. 1987). In Weiner et al. (1987), the authors found that good reasons usually tend to be external, uncontrollable (e.g., “My car broke down”) and internal, uncontrollable (e.g., “I could not come because of my broken legs”) to the actor, whereas bad reasons are usually internal controllable explanations (e.g., “I didn’t want to go”). The division of good/bad is in terms of whether the reason can trigger participants’ negative emotions (e.g., anger).

This distinction between “good” and “bad” reasons, based on their emotional impact, strongly resonates with pragmatic concerns, particularly the issue of (im)politeness (see Langlotz and Locher 2017 for a review on politeness and emotions). In the current study, we mainly draw on Spencer-Oatey’s (2002, 2005) rapport management theory, which conceptualizes politeness under the framework of interpersonal relations and provides a clear link between emotions and the interpersonal relations: rapport. Rapport can be simply interpreted as “harmonious social relations” and rapport management is thus “the use of language to promote, maintain, or threaten” this interpersonal relation (Spencer-Oatey 2008a, 2008b: 3). Spencer-Oatey (2002: 533–534) also defined “rapport sensitive incidents” as “incidents involving social interactions that they found to be particularly noticeable in some way, in terms of their relationships with the other person(s)”. This noticeable impact can be either positive or negative, which can be detected from the respondents’ positive or negative emotions: (1) happiness, pride, self-satisfaction and (2) annoyance, insult, embarrassment, humility. In her following work, she further summarized four groups of emotional reactions which can detect the respondents’ rapport management orientations: (1) joy (pleasure, pride); (2) surprise; (3) anger (annoyance, frustration, disgust); and (4) sadness (displeasure, guilt, embarrassment), and argued that these emotional reactions have an important impact on perceived rapport (Spencer-Oatey 2005: 134).

This model of rapport, grounded in emotional salience, offers a crucial lens through which attribution types gain pragmatic significance. Given that internal-uncontrollable and external-uncontrollable attributions have been shown to mitigate anger responses, whereas internal-controllable attributions tend to provoke stronger negative emotional reactions, their pragmatic implications can be reconsidered in light of rapport management theory (Spencer-Oatey 2002, 2005). Specifically, internal-uncontrollable and external-uncontrollable attributions may pragmatically function to soften the negative impact of refusal in rapport-sensitive incidents, whereas internal-controllable attributions are more likely to intensify that impact. Thus, internal-uncontrollable and external-uncontrollable attributions may contribute to a perceived rapport-maintenance outcome, defined as efforts to “maintain or protect harmonious relations between the interlocutors”, while internal-controllable attributions may lead to rapport neglect or even rapport challenge, which involves “a lack of concern or interest in the quality of relations” or an active attempt to “challenge or impair harmonious relations” (Spencer-Oatey 2008a, 2008b: 32).

The rapport-management function of attribution is likely to be realized via speakers’ management of accountability. Accountability refers to “the degree to which a person (or group of persons) is taken to be committed to or responsible for real-world consequences of their social actions and meanings” (Kádár and Haugh 2013: 119). As an important notion in post-2000 theorizations of politeness, it has been argued that our judgments about whether someone is polite or impolite are fundamentally based on how accountable they are for their social actions, considering what is expected in a given social situation (i.e., the moral order) (Haugh 2015). While all refusal explanations can be understood as forms of accountability practice, that is, attempts to justify a dispreferred social action in response to contextual expectations, not all explanations project the same level of accountability subject to interpersonal evaluation.

Internal- or external-uncontrollable attributions, for instance, present the refusal as constrained, involuntary, or circumstantially necessitated. These framings mitigate perceived accountability by minimizing the speaker’s apparent responsibility for the refusal. By doing so, they also reduce accountability stakes, the perceived interpersonal and moral consequences of being held accountable for a refusal in a given context, and thereby display an affiliative orientation that aligns with normative expectations in refusal discourse. In contrast, internal-controllable attributions make the speaker’s role in the decision more explicit, presenting the refusal as grounded in personal choice or preference. This openness to moral scrutiny leads to higher accountability, thereby elevating accountability stakes. For example, a refusal such as “I want to use it myself” offers no external rationale or constraint, thereby increasing the likelihood that the speaker will be perceived as inconsiderate or self-oriented, and that the refusee may feel devalued (i.e., face threat in Spencer-Oatey 2008a, 2008b). Such refusals thus orient to heightened accountability stakes and index a disaffiliative stance, which deviates from the affiliative orientation that is normatively expected in socially sensitive interactions. Through this lens, attributional strategies are not merely informational devices, but pragmatic resources for managing accountability and navigating rapport-sensitive refusals.

Recently, the attributional perspective has also been applied in pragmatics research. In a study comparing British and Australian requests, adopting Heider’s (1958) attributional perspective, Merrison et al. (2012) identified four types of accounts: institutional (relating to something within universities), health, employment, and personal. They also found that British students preferred to use personal accounts (i.e., internal factors), while Australian students often attributed their requests to external factors, such as institutional, medical, or employment factors. More recently, Ning et al. (2020) adopted two attributional factors: the locus of causality and controllability in their analysis of Chinese participants’ positive and negative evaluations of explanations for various social actions (e.g., requests, refusals, negotiations, persuasion, apologies). They identified four combinations of the two factors in their data and found that when explaining, participants prefer to use external uncontrollable reasons (43 %), followed by internal controllable reasons (30 %), internal uncontrollable reasons (13 %), and external controllable reasons (7 %). Moreover, the authors found that participants usually gave more positive evaluations (e.g., appropriate, acceptable) to external, uncontrollable reasons, whereas internal, controllable reasons received the most negative evaluations (e.g., inappropriate, hurtful), a finding consistent with Weiner et al. (1987).

While Merrison et al. (2012) and Ning et al. (2020) offer valuable insights into how people explain social actions, neither focuses specifically on refusals – a speech act with high rapport sensitivity. Whether their findings generalize to refusal contexts remains unclear. In particular, prior research has not examined how attributional framing interacts with contextual variables such as power and social distance. As a result, we know little about how speakers adjust attributional strategies to manage accountability across different relational configurations. In the next section, we introduce the two key contextual factors that constitute the focus of this study.

2.3 Contextual dimensions in interpersonal relations: power and social distance

Participant relations are widely recognized in pragmatics as being influenced by two key contextual variables: power (P) and social distance (D) (Spencer-Oatey and Žegarac 2017). P refers to “the ability to exercise control or exert influence on other people” (Spencer-Oatey and Žegarac 2017: 120), which is often “operationalized in terms of unequal role relations, such as teacher-student, employer-employee” (Spencer-Oatey 2008a, 2008b: 34). It is widely accepted that interacting with the people with a high-power status usually needs more politeness, linguistically and multimodally. D is defined as “a symmetric social dimension of similarity/difference”, often assessed by the frequency of interaction (Brown and Levinson 1987: 76). Findings on the effect of D are mixed, with some studies supporting a positive correlation between D and politeness (e.g., Brown and Levinson 1987; Leech 1983), while others report opposing evidence (e.g., Lim 1989 as cited in Lim and Bowers 1991; Wolfson 1988). Wolfson (1988) particularly emphasized that language does not become progressively more polite with greater social distance; instead, individuals tend to use the least politeness with those they are either very close to or barely acquainted with, and reserve the highest degree of politeness for relationships of moderate familiarity.

The impact of the two social variables is also observed in the speech act of refusals. Power is the most frequently investigated factor, and researchers have found that refusals to superiors typically involve more indirect strategies and supportive moves than those to inferiors (e.g., Kwon 2004; Lee 2013; Li 2022; Morkus 2014). However, research on the impact of social distance on refusals remains limited. Nevertheless, some insights can be gleaned from Bella’s (2014) data comparing refusal strategies across varying power and distance configurations. In her Table 1, a numerically higher frequency of direct and fewer indirect refusal strategies was observed in the P−/D+ scenario^[4] (i.e., unfamiliar neighbors) compared to the P−/D− scenario (i.e., friends). This pattern appears to support Lim’s (1989 as cited in Lim and Bowers 1991) and Wolfson’s (1988) discussion that politeness does not necessarily increase with greater social distance (e.g., addressing strangers).

Although empirical evidence suggests that social variables influence the production of refusals, much of the existing discussion has focused on broad categories such as directness or indirectness, rather than the specific refusal strategies employed. As Spencer-Oatey and Žegarac (2017: 119) argue, the interpretation of interpersonal relations, including politeness must be situated within two key relational dimensions: P and D. Given that different attributional types may carry different rapport implications, it is important to examine how speakers frame their explanations in relation to these contextual dynamics. Accordingly, our RQ2 investigates the distribution of attributional types across varying P/D configurations.

3.1 Participants

One hundred Chinese speakers participated in this study, involving 26 male and 74 female. Chinese speakers were selected as the target participants due to their extensive representation in existing politeness and refusal research. This established body of literature provides a robust foundation for comparing and connecting our findings with prior work. Furthermore, as a speech community often considered representative of East Asian cultural norms and encompassing the world’s largest population, investigating Chinese speakers offers valuable insights into pragmatic phenomena within this influential group. Participants were university students in Hebei with ages ranging from 18 to 21. They were recruited through advertisements spread in WeChat groups. They all signed consent forms and received explanatory documents before commencing the study. Upon completion of the study, participants received a small payment for their participation.

3.2 Instruments

This study used discourse completion tasks (DCTs) to elicit refusal explanations (see Appendix A). As the most widely used data elicitation method in speech act research, they offer the advantage in “elicit[ing] a large amount of data under controlled conditions from a large number of respondents in a short period of time” (Taguchi and Roever 2017: 110). However, we also acknowledge the limitations of DCT data. A major critique of DCTs is their elicited nature, which cannot accurately show the actual use of language in real life (Golato 2003), as well as their non-interactive nature, which provides limited insight into the dynamic, discursive features of real-world interactions (e.g., sequential unfolding) (Taguchi and Roever 2017).

Despite these critiques, House and Kádár (2021: 47) argue that “elicited data per se should not be dismissed as ‘invalid’ in general”. DCTs and naturalistic data represent different perspectives of pragmatics research. Naturalistic data primarily pays attention to accurately understanding human behaviors, such as research on ELF interactions (cf. Fang 2025), whereas DCTs mainly concentrate on uncovering “recurrent patterns and related norms behind such patterns” (House and Kádár 2021: 52). Since this study aims to unpack the refusers’ explanation pattern rather than examining the specific interactional resources that speakers use to organize explanations, DCTs remain suitable for our study. Moreover, while DCT data cannot capture the sequential or co-constructed nature of accountability negotiation, it does provide access to anticipated accountability concerns – what speakers orient to as plausible justifications when imagining a refusal scenario. Also, DCTs enable controlled elicitation across systematically varied social contexts, allowing us to examine patterns of attribution across multiple P/D configurations with a large number of participants. Importantly, we interpret our findings as reflecting participants’ projected orientations toward refusals (i.e., tendencies), rather than direct representations of real-world interaction.

Table 1 outlines the design of our DCT contexts, comprising four distinct contextual variations, each with two scenarios (i.e., eight scenarios in total). The DCT scenarios were designed to elicit refusals to requests involving imagined interlocutors who varied in terms of P and D. Participants were asked to write down their refusals using their usual and natural conversational style in Chinese. There was no time limit for them to complete the DCTs. They were neither notified that the study focuses on explanations nor instructed to include explanations in their responses. This design was intended to elicit more natural refusal responses. However, a potential risk of this design was that, without explicit instructions, participants might naturally omit explanations in favor of other refusal strategies.

To mitigate this potential risk and ensure the scenarios’ effectiveness in eliciting refusal explanations, we adopted a two-step approach. First, we selected and adapted scenarios from previous refusal literature that successfully elicited explanations. Subsequently, we conducted a pre-study survey on the DCT scenarios, presenting them to a separate group of 30 Chinese participants (who did not take part in the main study) and asking them to rate the likelihood of using explanations in the scenarios on a scale from 1 (definitely will not use) to 5 (definitely will use). The results indicated that the designed scenarios were effective in eliciting refusal explanations. The average ratings for each scenario were above 3 (neutral), with the mean ratings for Scenarios 1, 2, 3, and 4 all exceeding 4 (implying a high likelihood of using explanations), and the mean ratings for Scenarios 5, 6, 7, and 8 leaning toward 4, ranging from 3.1 to 3.7.

3.3 Data analysis methods

To answer RQ1 regarding participants’ thematic and attributional categories of refusal explanations, we coded the data. Since no existing coding scheme is available for the analysis of the content/themes of refusal explanations, this study adopted a bottom-up approach to process the DCT data. First, we identified all refusal explanations in the DCT dataset. Then, these refusal explanations were thematically summarized and categorized based on their content. Each explanation was analyzed to determine its thematic type.

After that, we further coded each refusal explanation based on two attributional dimensions: the locus of causality and controllability. Specifically, we examined whether participants attributed causes to internal (themselves) or external factors and whether they perceived/indicated control over the issues explained. The attributional coding of refusal explanations followed a five-category framework (see Table 2), comprising four types defined by locus of causality and controllability (cf. Ning et al. 2020), and an additional category for vague reasons. The four attributional types are internal-controllable (IC), internal-uncontrollable (IU), external-controllable (EC), and external uncontrollable (EU).

Our categories of IC, IU, and EU align with attribution theory (Weiner et al. 1987), but our operationalization of EC departs slightly. In Weiner’s model, EC typically refers to causes attributed to another agent (e.g., a partner or teacher) who is responsible for the outcome and could have acted differently to prevent it. This type of attribution is often associated with emotional responses such as anger or blame. In contrast, our EC category captures explanations where the cause is external, but the speaker retains some degree of control and signals a potential to act – for example, by suggesting they could reschedule, negotiate, or adjust their circumstances. This adaptation reflects the pragmatic orientation of our study: we are concerned with how speakers use language to construct social positioning and manage accountability. The EC category, as used here, enables us to analyze how individuals balance external justification with agentive stance-taking in pragmatic behaviors. In addition to the four attributional types, we also included a fifth category: vague reasons (VR) to account for refusal explanations in which the locus of causality or degree of controllability is unclear, ambiguous, or underspecified (e.g., I’m busy).

All thematic and attributional coding was conducted on the original Chinese responses, not the English translations. This approach ensured that coding decisions were grounded in participants’ actual linguistic choices and culturally embedded expressions. English translations were used only post hoc, for illustrative purposes in the reporting. To ensure reliability and consistency, an initial subset of 20 % of the refusal responses was independently coded by both researchers for both thematic and attributional categories, yielding an inter-coder agreement of 86.25 %. Following this initial reliability check, the remaining 80 % of the data was jointly coded, with any discrepancies resolved through discussion until full consensus was reached.

During the coding process, each refusal explanation was first assigned a concise content label capturing its core semantic meaning (serving as a subtheme). Semantically related labels/subthemes (e.g., “personal health/illness”, “personal inability”, “personal wellbeing”) were then further grouped into broader themes, such as Personal Capacity. This iterative, inductive process resulted in eight overarching refusal themes. A detailed discussion of subthemes is beyond the scope of this paper, but the full set of overarching themes and subthemes is provided in Appendix B. We acknowledge that the boundaries between themes were not always clear-cut. In cases of coding ambiguity, decisions were made through discussion and careful attention to the speaker’s primary framing. For example, the explanation 我正在准备一个很重要的考试 (‘I’m preparing for an important exam’) was coded as Commitment rather than Scheduling Conflict, as the speaker foregrounded the importance of the exam rather than a mere timing clash. Importantly, thematic coding was content-driven and conducted independently of attributional coding. This separation allowed us to explore how the same theme could be framed with different attributional meanings depending on the speaker’s interactional positioning.

In addition to reporting descriptive results, Generalized Linear Mixed Modeling (GLMM) was selected as the primary analytic approach to examine the relationship between contextual variables (i.e., power and social distance) and participants’ attributional patterns in refusal production. Notably, for the inferential statistical analyses, only participants (n = 63) who provided explanation responses in every scenario were retained for GLMM analysis. While this reduced the overall sample, it allowed us to preserve a fully crossed data structure appropriate for modeling context-dependent effects.

4.1 Thematic and attributional features of refusal explanations

As shown in Table 3, we have identified eight thematic categories from participants’ 777 tokens of refusal explanations, including (1) contextual constraints (n = 271), (2) scheduling conflicts (n = 181), (3) commitments (n = 116), (4) family (n = 89), (5) personal capacity (n = 79), (6) rights justifications (n = 25), (7) risk concerns (n = 11), and (8) interpersonal concerns (n = 5). All the examples were translated from our data.

As for the attributional features of the 777 refusal explanations, we identified four categories in the data: internal-controllable (IC), internal-uncontrollable (IU), external-uncontrollable (EU), and vague reasons (VR). As shown in Figure 1, IU attributions were the most frequently used (n = 279; 35.9 %), followed by EU (n = 216; 27.8 %). IC ranked third with 148 instances (19.0 %), while VR was the least frequent (n = 134; 17.2 %). Notably, no explanations were categorized as external-controllable (EC). Descriptively, participants tended to attribute their refusals to uncontrollable factors (IU + EU = 63.7 %) rather than controllable ones (IC = 19 %), and to internal rather than external causes (IU + IC = 54.9 % vs. EU = 27.8 %). Moreover, the data suggest that participants more often provided specific explanations with discernible attributional clues (82.8 %) than vague or underspecified ones (17.2 %).

Figure 1:

The distribution of attributional categories (note: values rounded; total may not equal 100 %).

Notably, certain refusal themes tend to co-occur with specific attributional types. As shown in Table 4, family-related and rights-based justifications are typically associated with EU attributions, while interpersonal concerns and personal capacity tend to exhibit internal attributional features. In contrast, themes such as contextual constraints, scheduling conflicts, and commitments display greater attributional variability, appearing across multiple types. This suggests that refusal themes do not map uniformly onto attribution categories, and the same theme may be produced differently depending on how speakers frame their reasons.

For example, the subtheme material unavailability under contextual constraints spans multiple attribution types, as in Scenario 6, where a classmate refuses to lend notes:

Although all three reasons stem from the same scenario, the participants construct distinct attributional explanations around the shared subtheme of material unavailability. P4 attributes the unavailability to her own academic priority (i.e., needing the notes for personal review), framing the constraint as internal and controllable. In contrast, P37 also locates the cause within herself, explaining that she did not take complete notes. However, as this is a past event she cannot now change, the reason is internal but uncontrollable. Unlike P4 and P37, P41 attributes the lack of notes to the teacher’s instructional style – an external factor beyond the speaker’s influence – thus rendering the refusal externally uncontrollable. These examples illustrate that even within the same theme and scenario, attributional types may vary depending on how participants frame the source and controllability of the reason.

Overall, this section reports eight frequently occurring themes and four attributional categories in our data. Building on the identification of thematic and attributional categories, the following section further analyzes the relationship between contextual features and the attributional types of refusal explanations.

4.2 Contexts and the attributional types of refusal explanations

This section presents the inferential statistical results regarding the relationship between contexts and attributional types. To enhance the interpretability of context effects while minimizing model bias due to unbalanced or structurally missing data, this section is based on data from participants (n = 63) who consistently provided reasons across all contexts.

Table 5 presents the raw frequencies of the four attributional types – IC, IU, EU, and VR – across four social contexts defined by power (P+ vs. P−) and social distance (D+ vs. D−), based on the 63 participants’ responses. The frequency patterns suggest that the use of attributional types varied considerably depending on power and social distance configurations. IC attributions were most frequently used in P−/D+ (n = 42), with less frequencies in P−/D− (n = 26) and P+/D− (n = 23). They were also rarely used in P+/D+ (n = 7). IU reasons were most common in P+/D− (n = 66) and P−/D− (n = 79), but far less used in P+/D+ (n = 31) and P−/D+ (n = 9). EU attributions showed the opposite trend: they were especially frequent in P+/D+ (n = 60) and P−/D+ (n = 67), with substantially fewer instances in P+/D− (n = 2) and P−/D− (n = 26). Vague reasons (VR) appeared most often in high-power contexts – P+/D− (n = 42) and P+/D+ (n = 37) – compared to P−/D− (n = 1) and P−/D+ (n = 9).

Moving to the inferential statistical analysis, the results confirm that participants were significantly more likely to use IC attributions in the P−/D+ context than in the P+/D+ context (p = 0.018). No significant differences were found in other pairwise contrasts. The finding indicated that IC usage was context-sensitive primarily when the interlocutor had a high degree of power and social distance, and suggested that participants were less inclined to assert personal priorities or choices in interactions with distant superiors (e.g., bosses) than with distant peers (e.g., strangers). This difference may reflect differing expectations of accountability and rapport management when refusing requests from individuals occupying distinct relational roles (see Section 5 for detailed discussion).

The results for IU attributions also revealed strong contextual effects. Participants were significantly less likely to invoke IU explanations in P−/D+ than in any of the other contexts. Specifically, compared to P−/D+, IU usage was significantly higher in P+/D− and P−/D− (p < 0.001 for both contexts), and was marginally significantly higher than P+/D+ (p = 0.053). Notably, in another distant context, P+/D+, IU usage was also significantly lower than in low-distance scenarios, specifically P+/D− (p = 0.012) and P−/D− (p = 0.001). This pattern suggests that IU attributions, such as reasons related to personal health or inability, were more likely to be used in familiar relationships, where self-disclosure of internal constraints may be socially acceptable or interactionally appropriate. It also demonstrates a main effect of social distance on IU attributions.

A markedly different pattern was observed for EU attributions. A significant main effect of social distance was found, with EU reasons used most frequently in socially distant contexts, regardless of power configuration. Participants cited EU significantly more in P+/D+ than in both P+/D− (p < 0.001) and P−/D− (p = 0.005), and also used it more in P−/D+ than in those same two conditions (p < 0.001 for P+/D−; p = 0.002 for P−/D−). In these distant interactions, speakers often invoked reasons beyond their control that were external in origin, such as obligations to family members, institutional rules, or third-party interference.

Finally, VRs were also shaped by context in systematic ways. A clear main effect of power emerged for VRs: speakers were significantly more likely to use vague explanations in P+/D+ than in P−/D+ (p = 0.005) and P−/D− (p = 0.001), and likewise more likely to do so in P+/D− than in P−/D+ (p = 0.004) and P−/D− (p = 0.001). This pattern indicates that VRs were most frequently used in high-power contexts, regardless of social distance.

In sum, the GLMM analyses confirm that the selection of attributional types in refusal explanations was significantly shaped by social context. Each attribution type followed a distinct distributional profile: internal reasons were used most in symmetrical or close interactions; external and vague reasons predominated in hierarchically or socially distant interactions. These patterns demonstrate that participants’ attributional framing is not only semantically strategic, but also deeply responsive to the relational configurations in which refusals occur.

5.1 Thematic and attributional types of refusal explanations (RQ1)

Thematic analysis of the Chinese participants’ refusal explanations revealed eight recurring themes: contextual constraints, scheduling conflicts, commitments, family, personal capacity, rights justifications, risk concerns, and interpersonal concerns. While our participants were Chinese speakers, the thematic categories identified, such as contextual constraints, commitments, scheduling conflicts, and family, etc., echo those displayed in previous refusal studies across a range of linguacultural contexts. This suggests that certain explanation themes may be recurrently salient across different speech communities. For instance, in Liao and Bresnahan (1996), their Chinese and American participants gave refusal explanations that fell under several of our identified themes, including scheduling conflicts (e.g., M2m,^[6] M3m), (academic) commitments (e.g., A28m, A30m), and contextual constraints such as material unavailability (e.g., M10m, A12m). Similarly, Kwon (2004) reported that Korean participants often cited their parents or in-laws and physical discomfort, aligning with our “family” themes. Moreover, our categories of financial constraints and scheduling conflicts parallel Siebold and Busch’s (2015) themes of lack of money and time, respectively. Hence, a key contribution of this study lies in rendering these often-implicit thematic patterns more analytically visible. By organizing refusal explanations into a systematic and replicable table (i.e., our Table 3/Appendix B), the study offers a transparent framework that can support future investigations into the content and variation of refusal discourse across contexts and linguacultures.

The attributional analysis of the data exhibited four attributional types in refusal explanations: internal-controllable (IC), internal uncontrollable (IU), external-uncontrollable (EU), and vague reasons (VR). Different from Ning et al. (2020), who found that EU was the most frequent attribution type followed by IC and then IU across a variety of speech acts (e.g., requests, refusals, persuasion, negotiation, and apology), our results reveal a distinct pattern within the refusal context: IU emerges as the most frequently used attribution, followed by EU, while IC is used the least.

This divergence likely results from the unique accountability demands of refusal as a rapport-sensitive social action. Unlike many other speech acts, refusals often carry a higher risk of rapport challenge (Spencer-Oatey 2008a, 2008b), especially when directed toward individuals in asymmetrical or socially significant relationships. In such contexts, speakers are under pressure not only to explain their disalignment, but also to do so in a way that minimizes personal responsibility for preventing moral judgement. According to Weiner et al. (1987), IU attributions, such as citing physical limitations, illness, or emotional states, enable speakers to acknowledge internal causality while simultaneously framing it as beyond their volitional control. This allows refusers to maintain rapport via reducing the likelihood of eliciting anger – a key emotional indicator of rapport challenge (Spencer-Oatey 2002). Consequently, IU emerges as a socially defensible and face-sensitive form of account, foregrounding the speaker’s “no-fault quality” (Heritage 1984: 271) and claiming interactional accountability for how their refusal is understood appropriately within politeness frameworks (Kádár and Haugh 2013). In contrast, IC attributions, where the refusal is traced to the speaker’s volitional decisions, entail a stronger claim to personal responsibility without reducing the accountability stakes in refusals. Hence, they are potentially rapport-challenging, and as our findings suggest, they are less common in our data.

Notably, EC attributions were absent from our data – a finding that is both empirically striking and pragmatically expected. EC attributions in this study refer to refusal reasons where the cause lies outside the speaker but remains subject to their control. In the context of refusals, such attributions create a strategic dilemma: if the speaker acknowledges control over the external constraint but still chooses to refuse, the implication is one of unwillingness, not incapacity. This undermines the speaker’s attempt to appear morally exonerated. As a result, EC attributions risk being interpreted by interlocutors as deliberate withholding of help, which can threaten the interlocutor’s face and damage rapport. The underlying message may be taken as “I could help, but I won’t”, which sharply contrasts with the more affiliative framing offered by uncontrollable attributions (e.g., EU, IU). In this sense, EC reasons may be pragmatically incoherent with the broader social goals of refusal explanations, namely, to present the speaker as accountable yet legitimized and appropriate. Hence, it is understandable why they did not occur in our data.

Furthermore, our study found that vague reasons for refusals were uncommon, making up only 17 % of the data. This finding aligns with previous research indicating that Chinese speakers often provide detailed explanations when refusing (e.g., Chang 2011; Liao and Bresnahan 1996). This likely reflects Chinese participants’ preference for politeness via elaborated accounts. As Rintell and Mitchell (1989) suggest, more elaborate speech acts are often perceived as more polite by the recipient.

5.2 Context variations in attributional types (RQ2)

Our findings show systematic patterns between social context (power and social distance) and attributional framing, even though the interlocutors and contexts are imagined. This suggests that participants are orienting to shared social norms and accountability expectations, even in non-interactive conditions. Specifically, we found that IC attributions were significantly more common in P–/D+ context (e.g., strangers, unfamiliar neighbors). In contrast, such attributions were markedly less frequent in P+/D+ contexts, such as those involving distant authority figures (e.g., bosses). This distribution suggests that participants seem to pay limited attention to the accountability stakes in the P−/D+ contexts, where rapport obligations are minimal and moral evaluation is less consequential. As Wolfson (1988) observed, individuals tend to speak least politely with those they barely know. Accordingly, participants may have felt more at liberty to present refusals as personally volitional without fearing negative emotional reactions (Ning et al. 2020; Weiner et al. 1987) or the need to protect rapport (Spencer-Oatey 2002).

However, the frequent use of IC attributions in P–/D+ contexts does not suggest that participants uniformly adopt rapport-neglect orientations. Our data also show a significant use of EU attributions in this context. This co-occurrence highlights the complexity of social distance as a contextual variable in pragmatic research, aligning with previous mixed findings regarding the relationship between high social distance and politeness (see Section 2.3). On the one hand, some participants’ preference for EU attributions, which are less likely to elicit negative evaluations (Ning et al. 2020; Weiner et al. 1987), may reflect intentional politeness-related efforts under conditions of high social distance (Brown and Levinson 1987; Leech 1983), framing their refusals in a socially acceptable and morally defensible manner. On the other hand, high distance may also reduce the perceived need for rapport management (Lim 1989, as cited in Lim and Bowers 1991; Wolfson 1988), thereby encouraging some participants to adopt more self-oriented attributions such as IC. What drives participants’ divergent orientations toward high social distance may require further investigation, possibly through analysis of their metapragmatic comments on their task performance. However, any possible findings would be the subject of another research.

In addition to P–/D+ contexts, EU attributions are also frequently employed in P+/D+ situations (e.g., with bosses). This tendency can be interpreted as a strategic attempt to mitigate accountability. When rejecting superiors, individuals are generally expected not only to be held accountable for their actions, but also to frame the refusal as both appropriate and morally justifiable. When emphasizing unavoidability due to external constraints, participants diffuse potential blame and show affiliativeness to the higher-status interlocutor. In particular, EU reasons in our data often invoke the health issues of elderly parents and the participants’ caregiving responsibilities. These justifications serve to both soften the speaker’s perceived accountability and reinforce the moral legitimacy of the refusal, particularly in the Chinese sociocultural context, where filial piety is central to the moral order and highly valued (Ivanhoe 2004).

In close interactions (P–/D–, P+/D–), participants also demonstrated their concerns for mitigating accountability and maintaining rapport, as evidenced by their significantly greater use of IU attributions. Like EU, IU enables speakers to frame their refusals as legitimate and non-blameworthy, thereby increasing the likelihood of positive evaluations (Weiner et al. 1987) and minimizing relational disruption (Spencer-Oatey 2002). The frequent deployment of IU in low social distance contexts can be understood in light of Lim’s (1989, cited in Lim and Bowers 1991) finding that greater politeness-related efforts are often made with intimates than with acquaintances. Moreover, Lim (1990) argues that high-imposition occasions – even among close interlocutors – typically invite more politeness-related work. Although the present study did not explicitly manipulate the rank of imposition, both low-distance scenarios carry a relatively high interpersonal risk for the refuser, which likely intensified participants’ awareness of potential rapport threats. These may help explain the increased use of IU as a rapport-maintaining strategy in this context. Notably, participants usually attribute their uncontrollability to internal vulnerabilities (e.g., health issues, perceived inabilities) in these contexts. This suggests a potential link between self-disclosure and familiarity between interlocutors. This interpretation also aligns with the psychological observation of interpersonal relationships: more personal and private matters are only disclosed to significant others (e.g., Altman and Taylor 1973).

Different from IU, VRs seem to be more sensitive to power rather than social distance, frequently used in asymmetric relationships including P+/D+ and P+/D–. This observation is consistent with Liao and Bresnahan (1996), who also observed that vague reasons are frequently used by their participants to address superiors. The use of VRs in such contexts suggests an effort to refuse while minimizing personal exposure, thus avoiding potential moral scrutiny from superiors. Although the rapport-maintaining function of VRs may be weaker compared to lengthy, specified explanations (Rintell and Mitchell 1989), the complete refusals in our data typically include additional strategies, such as apologies and alternatives, that may help compensate for this weakness when addressing superiors.

Overall, this study contributes to an in-depth understanding of refusal explanations beyond the traditional focus on their indirect forms and content (e.g., specificity and topics). By integrating attribution theory, we show that refusal explanations, traditionally often viewed as a monolithic category based on Beebe et al. (1990), are internally differentiated based on how speakers assign causal responsibility and perceived control. These attributional choices are systematically shaped by relational context variables, particularly power and social distance. Participants actively select attributional types that align with the social demands of the context, thereby managing emotional consequences and mitigating potential rapport threats.