In pragmatics research, the term speech acts refers to “functions performed through language” or “doing things with words” ( Ishihara & Cohen, 2022, p. 11) and is among the most well-researched topics in pragmatics ( Taguchi & Roever, 2017; Taguchi et al., 2016; Tatsuki & Houck, 2010). There are many kinds of speech acts, such as greeting, inviting, requesting, refusing, apologizing, complaining, complimenting, and thanking ( Ishihara & Cohen, 2022, p. 11). While requests are the most studied speech act, some types ( e.g., teasing and criticizing) have not been examined sufficiently ( Cohen, 2020). It is also known that certain speech acts are strongly influenced by the speaker’s native language ( e.g., Cohen, 2020; Yates & Springall, 2010). For example, Japanese speakers tend to show disfluency when producing apologies and refusals as a means of politeness toward interlocutors ( Cohen, 2020).

Several assessments have been used to examine the production of speech acts. The most traditional and frequently used task is a discourse completion task ( Taguchi & Roever, 2017). In a DCT, a certain context is given to participants, who are then asked what they would say in that situation. A DCT can involve a free response or a cued response multiple-choice ( Hudson et al., 1995). While several researchers have criticized written DCTs because they lack authenticity and it is not clear if the participant would really speak in a particular way in the real situation ( e.g., Taguchi & Roever, 2017), some researchers use them in addition to other assessments ( e.g., Gilmore, 2011). DCTs have strengths in terms of their practicality and ease of use in data collection ( Cohen, 2020). Recent data collection methods include oral DCTs, role play, naturalistic data gathering such as through recordings ( Cohen, 2020), and the use of virtual reality ( Taguchi, 2022).

The significance of L2 pragmatics has been recognized in the era of globalization. Specifically, the number of L2 pragmatics studies in the context of English as a lingua Franca (ELF) has been increasing, and this new type of work has transformed and expanded the traditional L2 pragmatics research in several ways, such as emphasizing the role of L2 speakers in negotiations of meaning when they face communication difficulties ( Culpeper et al., 2017). However, compared with cross-cultural pragmatic studies, there has been scarce research that has examined the connection between L2 pragmatic and intercultural competence ( e.g., Taguchi et al., 2016). Intercultural competence is important in contexts in which cultural differences are found and is attributed by individual factors including “personal qualities, attitudes, knowledge, and skills” ( Taguchi et al., 2016). Recent empirical studies have verified the positive effect of studying abroad on L2 pragmatic learning ( e.g., Alcón-Soler, 2015; Cohen & Shively, 2007; Taguchi et al., 2016), which indicates that being exposed to a culture plays an important role in developing L2 pragmatic skills. In fact, choosing the appropriate phrases to use in a particular social situation requires a great deal of cultural understanding, which reveals the speaker’s understanding of their membership within their respective speech communities. Bardovi-Harlig (2012) revealed that while these phrases are not always formulaic, researching the use of formulaic language helps to emphasize the importance of context related to the pragmatic nature of such formulaic social interactions.

Taguchi (2022) argued that due to English’s status as a dominant language utilized in intercultural contexts by nonnative speakers, native-speaker norms cannot function as the only reference for denoting the sufficiency of nonnative speakers’ English. Specifically in the context of English for Occupational Purposes (EOP), the native speaker’s norm is not as crucial as that of an expert ( Tarone, 2008).

Recent EOP studies have investigated pragmatics in the following disciplines: medical ( e.g., Dahm et al., 2022), business ( e.g., Kaur & Birlik, 2021; Park et al., 2021), aviation ( e.g., Ishihara & Prado, 2021), and the professional kitchen ( Pang, 2019). Kaur and Birlik (2021) investigated the pragmatic strategies that are used in business meetings at a multicultural corporation. The strategy scrutinized in this study was explaining that either followed an implicit request or was done without such a request. Data were collected from six business meetings; although the personal profile of each participant was not clear, the participants’ native languages were Malay, a Chinese dialect, Tamil, and German. The major finding of this study was that the participants tended to add some background information so that the attendees of the meeting would establish shared knowledge. Park et al. (2021) explored how requests are made in a business context, and compared 50 Korean and 50 American office professionals’ email writing. Data was collected through a written DCT, and the collected emails were analyzed focusing on moves as well as lexical and syntactic complexity. They revealed some cultural differences between the Korean and American professional groups. First, members of the Korean group provided more self-introduction than those of the American group. Second, the American group used specific strategies, such as complimenting and offering compensation, more often than the Korean group. Third, the subject lines were longer and more detailed in the Korean group than in the American group. Park et al. explained that these differences reflect sociocultural differences between Korea and America.

The production of requests and directives was examined in a professional kitchen in Pang (2019). The study focused on interactions, and the data was collected using field notes and participant observation, and the participants’ workplace communication was also recorded. The participants comprised eight trainees and 55 kitchen workers. Pang (2019) revealed that the strict hierarchy in the kitchen created an environment in which a clear distinction in power status allowed directives to be issued by workers with higher authority.

Ishihara and Prado (2021) scrutinized the pragmatics in radiotelephony communications between pilots and air controllers. Their focus was on negotiation of meaning, and data were extracted from the Radiotelephony Plain English Corpus (RTPEC) ( Prado & Tosqui-Lucks, 2019). The authors claimed that aviation English and particularly radiotelephony communications have some similarities and differences with general ELF.

While the number of L2 pragmatic studies in the context of EOP has been increasing, so too has the importance of investigating a wide variety of populations ( Pang, 2019; Park et al., 2021) and using multiple types of strategy ( Kaur & Birlik, 2021).

Although the pragmatic differences between Japanese and English have been mentioned in previous studies ( e.g., Cohen, 2020; Yates & Springall, 2010), compared with cross-cultural pragmatic studies that have compared English and Japanese languages ( e.g., Fukushima, 1996) and L2 pragmatic development by Japanese EFL university students ( e.g., Matsumura, 2003; Takahashi, 2005, 2015; Takahashi & Beebe, 1987), few empirical studies have examined Japanese workers who have worked overseas while focusing on L2 speech act production.

Kubota and Takeda (2021) conducted political discourse analysis on governmental documents and interviews with corporate workers. From their interviews with Japanese transnational workers in Asia, it was found that Japanese corporate workers did not believe the connection between English test scores and job satisfaction ( Kubota & Takeda, 2021). It is important to recognize that obtaining good scores from the Test of English for International Communication (TOEIC) does not guarantee the test taker’s success in his/her business, although not achieving native command tends to be regarded as not sufficiently proficient in Japan ( e.g., D’Angelo, 2018).

According to the statistics reported by the Ministry of Foreign Affairs of Japan as of October 1, 2021, there are as many as 807,238 Japanese citizens abroad. These individuals were reported as staying in a foreign country for longer than three months but planning to return to Japan in the future ( Ministry of Foreign Affairs of Japan, 2022a). Also, the Ministry of Foreign Affairs of Japan (2022b) revealed that 77,551 companies operate overseas and that the number of business operations in Asia ( n=53,431) is much higher than that in North America ( n=9,827) and that in Europe ( n=8,300). However, how Japanese workers use language pragmatically in an intercultural workplace has been under researched.

Considering the background discussed so far, the purpose of the current study is to clarify (a) challenging aspects in L2 pragmatics for Japanese workers and (b) individual difference factors ( e.g., self-reported English proficiency, overseas work experiences) that influence speech act production. We have established the following research questions:

RQ1: What influences the pragmatic score more: frequency of strategy use or individual difference factors?

RQ2: What types of pragmatic strategy lead to higher pragmatic scores?

RQ3: What perceptions do Japanese workers who have been employed in foreign countries have toward the intercultural workplace communication?

This study received ethics approval from the University of Tsukuba Faculty of Humanities and Social Sciences Ethics Committee on January 14, 2021, with the approval number 2020-10. Potential participants read the explanations about this study and the participation instructions on a Japanese crowdsourcing service called CrowdWorks. It was impossible to obtain written consent from the participants because participants do not disclose their names when they accept tasks on CrowdWorks. Therefore, we clearly outlined that participating in this study would indicate that the participants had read the explanation, which stated that the data would be analyzed anonymously and used for academic publications, and had agreed to join on their own accord. The study explanation also clearly stated that the purpose of this study was to investigate workplace communication produced by Japanese workers and that the task would require potential participants to write down English expressions. When collecting data from an online survey, it was necessary to consider the type of question to include in order to avoid careless responses from the participants. According to Brühlmann et al. (2020), careless responses are a form of invalid responding, such as seeking social desirability and faking responses. It was also mentioned that careless responding is a phenomenon typically due to factors such as extrinsic motivation, length of survey, social contact, environmental distraction, as well as motivation and interest, which can negatively affect data collection in online survey responses. Furthermore, open-ended questions have been found to receive fewer careless responses ( Brühlmann et al., 2020). Only those who agreed to take part in this study including such open-ended questions clicked on the button to join this study; thus, it was considered that having anonymized responses did not distort the scientific data. The University of Tsukuba Faculty of Humanities and Social Sciences Ethics Committee approved this form of implied consent acquisition.

We recruited 100 native Japanese speakers through CrowdWorks. The sample size was determined considering the two facts. The first one is that some pragmatic studies using written DCT had similar sample sizes ( e.g., Jebahi, 2011; Park et al., 2021). The second one is that CrowdWorks can send our call to 100 potential participants who have specific qualifications as the optional service. We wished to reach out to those who have English skills, so having 100 participants seemed reasonable for our recruitment possibility. Due to problems with linking the crowdsourcing service and the questionnaire platform, Microsoft Forms, eight people were excluded from the data analysis. Therefore, 92 native Japanese speakers were included in the data analysis. Table 1 shows the demographic information of the participants.

Nearly one-third of the participants ( n=34), including two participants who lived in English-speaking countries over 20 years, did not have any English test scores. Almost half ( n=41) of the participants reported their TOEIC scores, and their mean score was 818.54 ( SD=120.16), which falls in B2 of the Common European Framework of Reference for Languages (CEFR) proficiency levels ( ETS Global, 2021). However, their scores ranged from 430 to 990, which were equivalent to CEFR A2 to C1. Also, nearly one-fourth ( n=24) of the participants had Test in Practical English Proficiency (EIKEN) grades ranging from Grades one to four. The EIKEN grades, except Grade four, can be mapped with CEFR as follows ( Eiken Foundation of Japan, 2022): Grade one is mapped to B2 and C1, Grade pre-one is mapped to B1 and B2, Grade two is mapped to A2 and B1, Grade pre-two is mapped to A1 and A2, and Grade three is mapped to A1. It should be noted that three participants have lived in America or the UK for more than 20 years, and they declared that they had native-command English proficiency. Based on these test scores/grades and their self-reported English proficiency bands, the proficiency levels of the participants were estimated from A1 to C1.

The materials included a demographic questionnaire, survey questions, and a written DCT, which can be found in Open Science Framework ( Hijikata et al., 2022). The demographic questionnaire and survey questions were included in Part one, followed by Part two, a written DCT. The contents of the questionnaire included items on the participants’ age, gender, working status, languages used at work, and overseas work history. The survey questions featured items about participants’ experiences and challenges in working with colleagues with different linguistic and cultural backgrounds, what they do to overcome such challenges, and what they believe are essential to work abroad.

We developed and conducted a written DCT based on one of the authors’ own experiences working abroad. Twelve situations that could occur in workplaces were written in Japanese. Here is one example:

[Situation 1] You are working as a customer service representative of a research company. You are calling your client for him/her to answer an online survey. This survey will help improve the sales representatives’ performance and will take five minutes of their time. What would you say to your client? [Request]

The DCT included speech acts of requesting, advice giving, apologizing, and refusals. Each speech act featured three situations, and the presentation order was randomized using Excel for Microsoft 365.

All data were collected through Microsoft Forms, which was linked to CrowdWorks. When recruiting participants, the advertisement called for participants who regularly used English at work or who had worked abroad. To reach people who use English at work, we added an option to CrowdWorks that allowed us to send our research advertisement to 100 people who had registered English as part of their skillset. The data collection started on January 11 and ended on January 12 in 2022.

This study consisted of two parts. The first was a survey with 19 questions to collect the participants’ demographic data and information about their beliefs as to what is important for Japanese workers working overseas and/or with colleagues from different countries/cultures. The second part was a DCT with 12 questions. The participants were asked to write “I don’t know” in Japanese when they were not sure how to respond. They were required to type what they would say in each situation in English and judge how often they encountered such a situation using a five-point scale (“Never,” “Once in 2–3 months,” “Once a month,” “Once in a couple of weeks,” and “More than 3 times a week”). The participants were instructed not to search for information on the Internet nor consult expressions with a dictionary while they engaged in the DCT. When they completed the study, they were given a completion code on Microsoft Forms. They were expected to input the completion code into CrowdWorks to receive their monetary reward. They received JPY 860 as compensation for their time.

Scores of speech acts

The speech acts produced were graded using rating scales contextualized by three perspectives: whether the expressions were appropriate in each context, if communication was clear, and whether each response was grammatically accurate as a whole ( Taguchi et al., 2016). Possible scores were 1–6.

One-fourth of the data was graded by two authors, one native English speaker who has taught English in Japan for over six years, RDR, and one native Japanese speaker, who has taught English for more than 20 years, MU. The interrater reliability was checked for each question using R (version 4.2.1) with the vcd package. The initial reliability kappa (weighted) was 0.65. One question had lower interrater reliability, so the two raters discussed the disagreement; as a result, the interrater reliability improved to 0.74. Then, the remaining three quarters of the data were graded by the native English speaker, RDR only.

Strategies

We adapted separate frameworks for each speech act with reference to the frameworks used in previous studies ( Blum-Kulka et al., 1989; Ishihara & Cohen, 2022; Taguchi, 2022). Strategies in requests were further divided into main strategies that directly convey the act, and modifications that support the main strategies. Ten categories in the main strategies were adapted from Taguchi (2022), and we created one new category, “Japanese-like” because some responses were highly influenced by the Japanese language. We show the mapping between categories adapted from Taguchi (2022) and thus indicated in the quotation marks and sample responses in our study as follows. (1) “ Imperative” includes imperative sentences with and without please ( e.g., Please answer my questions/Please cooperate). (2) “ Performative” includes statements which require a listener to act something or statements with the verb request ( e.g., I’m requesting a questionnaire …). (3) “ Obligation” indicates a duty to do something, so the use of structures and/or modal verbs such as should and must fall in this category (e.g., It is your duty to … / You should cooperate). (4) “ Want statement” indicates speakers’ willingness to do something, so this category contains verbs such as want, wish, and need, and phrases such as would like to ( e.g., I want to know … /I would like to ask you …). Different from (2) performatives, the want statement does not force the listener to do something. (5) “ Direct question” asks a simple question which does not get permission to act something ( e.g., Do you have time to answer … ?). (6) “ Query preparatory” is in a question form to ask a hearer to ask a favor ( e.g., Could you cooperate … ?/Can you give me 5 minutes?). (7) “ Permissions” seeks permission to do something ( e.g., May I ask you … ?/Can I … ?). (8) “ Suggestions” shows giving a suggestion to a hearer ( e.g., Why don’t you answer … ?). (9) “ Mitigated expressions” represent embedded clauses ( e.g., We appreciate if you can …). (10) “ Hint” does not clearly ask a hearer to do something for the hearer but implies that such action is desirable ( e.g., This survey is designed to improve … and takes about 5 minutes to complete). (11) Japanese-like is an expression influenced by Japanese and does not ask whether a hearer accepts the request or not ( e.g., Thank you for your cooperation for 5 minutes). The categories (1)–(5) are direct expressions, while the categories (6)–(11) are indirect questions. Responses which did not apply to any of these were marked as (12) NA. Furthermore, these main strategies were supported by the following modifications, which were also adapted from Taguchi (2022): (a) “ Hedging,” (b) “ Reason,” (c) “ Apology,” (d) “ Preparator,” (e) “ Confirmation,” (f) “ Appreciation,” (g) “ Minimizer,” (h) “ External ‘please’,” (i) “ Question,” (j) “ Promise,” (k) “ Positive Comment,” (l) “ Alternative,” and (m) “ Attention Getter.”

Next, strategies in apology were coded based on the main strategies only. Five categories with the quotation marks were adapted from Ishihara and Cohen (2022), and we created one new category, Japanese-like. Thus, all main strategy categories related to apology are (1) “ apology,” ( e.g., We apologize/Sorry) , (2) “ responsibility,” ( e.g., It is our fault), (3) “ explanation,” which states why the trouble happened ( e.g., We seem to have forgotten to check …) , (4) “ compensation,” which offers compensation ( e.g., We will offer you complimentary drinks …), (5) “ promise for the future,” ( e.g., this won’t happen again), and (6) Japanese-like, with the discourse strongly affected by the Japanese style.

Regarding refusals, the main strategies included 10 categories outlined by Taguchi (2022), and we created one new category, Japanese-like, because some responses were highly influenced by the Japanese language. Thus, all main strategy categories adapted from Taguchi (2022) as shown in the quotation marks and sample responses in our study are as follows: (1) “ non-performative ‘no’,” ( e.g., It is not refundable) (2) “ negative willingness/ability” ( e.g., We can’t …), (3) “ mitigated refusal,” (e.g., I’m afraid we are unable to …) (4) “ indefinite reply,” which is an evasive response, (5) “ excuse,” which states the reason why the request is not acceptable (6) “ wish,” (7) “apology,” which turns down the request with apology expressions, (8) “ avoidance,” which avoids declining the request clearly, (9) “alternative,” which suggest something else instead of accepting the request, (10) “ promise,” which implies that the request will be accepted in the future ( e.g., We will have such instructors in the future), and (11) Japanese-like. Furthermore, these main strategies were supposed by the following modifications, which were also adapted from Taguchi (2022): (a) “positive comment,” (b) “well-wishing,” (c) “appreciation,” (d) “apology,” (e) “reason,” (f) “hedging,” (g) “confirmation,” (h) “external ‘please’”, and (i) “attention getter.”

Finally, advice-giving was analyzed from the main strategies framework only. These main strategies included three categories by Ishihara and Cohen (2022): (1) “ direct,” (2) “ softened,” and (3) “ indirect.” Their labeling and examples in our study are shown as follows. (1) “ Direct” is straightforward, and the person who gives advice use imperatives or uses modals such as should and had better ( e.g., Please consult). (3) “ Indirect” leaves some room for not accepting the advice by using mild expressions ( e.g., You might want to listen to others’ opinion). (2) “ Softened” is even less direct (e.g., We are the team, and we are here to help you).

The strategies were coded by classifying speech act types by the authors: One coder, YH, had a doctoral degree in applied linguistics and had taught general and academic English at tertiary level for over 12 years and occupational English in the US for four years. She coded all of the main strategies, while RDR and MU coded half of the responses, respectively. The modification was coded by YH. The overall agreement rate between YH and RDR/MU was 77%. The disagreement was solved through discussion.

To calculate the overall scores and frequency of each strategy, we ran separate models. Generalized linear mixed-effects (GLMM) models using R (version 4.2.1; R Core Team, 2022) and R Studio (version 2022.07.1) with the lme4 package were used to examine RQ1. Our focus was the relationship between the pragmatics scores and the types of strategy, so we counted the number of frequencies and performed Fisher’s exact tests in a statistical analysis program called js-STAR ( Tanaka & Nakano, 2022) and R studio to answer RQ2. RQ3 was mainly analyzed qualitatively, but Fisher’s exact test was also used. We included data even if no English expression was written for each question because we also counted the frequency of “non-applicable” (NA) in the data analysis. However, three out of 12 questions from one participant, who had lived and worked abroad for over a year, were excluded from the data analysis because he had to complete the experiment in 60 minutes to get compensation from CrowdWorks. In his case, coding his blank cell as NA could distort the relationship between scores and strategies.

The participants’ individual factors were also sum-coded as follows: English use at work (Yes: 0.5, No: −0.5), work experiences in an English-speaking country (Yes: 0.5, No: −0.5), self-reported English proficiency (Native: 1, Advanced: 2, Upper intermediate: 3, Intermediate: 4, Beginner or lower intermediate: 5, Limited: 6).

In this section, we report on our analysis of which factors influenced the pragmatic scores. As Table 2 and Figure 1 indicate, the mean scores were consistent for the speech act types, although refusals was the lowest and request was the highest.

Following Taguchi’s (2022) analytical framework, we counted the number of strategies for each speech act. There were generally three requesting strategies, and the generalized linear mixed model (GLMM) using the Poisson distribution was fit by maximum likelihood (Laplace approximation). The best-fitting model included the following fixed effects: the number of main strategies, the number of modifications, self-reported English proficiency, and experience working overseas. Random effects included both participants and items, and a random slope of English use at work was applied to the participant data. The best-fitting model was y ∼ nmain + nmodi + prof + exp + ( 1 + use | ID ) + ( 1 | t ) , data = datf , family = “poisson.”

It was shown that the number of main strategies ( β=0.22, SE=0.07, z=3.33, p<0.001) and that of modifications ( β=0.11, SE=0.03, z=3.87, p<0.001) significantly improved the pragmatic score. However, neither the experience of working overseas nor the self-reported English proficiency significantly affected the score.

Apology was assessed using three questions, and the pragmatic score was modeled using the GLMM. Since this speech act did not have a framework of modifications, the fixed effects in the best-fitting model contained the number of main strategies, self-reported English proficiency, and experience working overseas. Random effects included both participants and items, and a random slope of English use at work was applied to the participant data. The best-fitting model was y ∼ nmain + prof + exp + 1 + use ID + 1 t , data = datf , family = “poisson.”

It was shown that an increase in the number of main strategies significantly increased the pragmatic score ( β=0.25, SE=0.05, z=5.44, p<0.001). Experience working overseas did not significantly affect the score. The advanced and beginner/lower intermediate groups showed a significant difference in English proficiency.

Refusals included frameworks of main strategies and modifications, so the frequencies of these strategies were counted and included in the GLMM. The best-fitting model was quite similar to that of the model for requests: y ∼ nmain + nmodi + prof + exp + 1 + use ID + 1 t , data = datf , family = “poisson.”

An increase in the number of main strategies led to higher pragmatic scores ( β=0.53, SE=0.10, z=5.36, p<0.001), and this was also the case for the number of modifications ( β=0.11, SE=0.04, z=3.00, p=0.003). In this model, those who had experience working in a foreign country had significantly higher scores than those without such experiences ( β=0.16, SE=0.08, z= 2.03, p=0.004); however, English proficiency was not a significant fixed effect.

Finally, advice giving strategies were analyzed using the framework of the main strategies only. The best-fitting model was quite similar to that for apology: y ∼ nmain + prof + exp + 1 + use ID + 1 t , data = datf , family = “poisson.”

The number of main strategies was the only significant factor that improved the pragmatic scores ( β=0.83, SE=0.12, z=6.67, p<0.001), while other fixed effects were not significant.

In sum, the more main strategies speakers use, the more appropriately the speech acts sound in English. This tendency was confirmed regardless of the speech act. Only refusals showed a significant fixed effect of overseas work experience. This means that Japanese speakers learn how to appropriately refuse someone’s request in English by living and working abroad.

In this section, we scrutinize the relationship between the pragmatic scores and the strategies used in each question. Since the number of categories differs depending on the speech act, we analyzed the data for each speech act.

Requests

Requesting strategies outlined in Q1, Q2, and Q6 were coded from the main strategies and modifications. Table 3 and Figure 2 show the relationship between pragmatic scores and the main strategies used in requests. Furthermore, Table 4 and Figure 3 represent the relationship between pragmatic scores and modifications in requests made.

The Fisher’s exact test results were significant ( p=0.0005 [main strategies]/ p=0.0005 [modifications]), indicating that each score group used different strategies. The post hoc analysis with Bonferroni’s adjustment showed that, as the main strategies, those who scored 6 used Strategy 9 (mitigated expressions) more frequently more than the expected value ( z=4.58, adjusted p=0). Those who scored 3 tended to use Strategy 1 (imperative) more than the expected value ( z=3.61, adjusted p=0.018). Those who scored 1 did not use any strategies and were categorized as NA, more than the expected value ( z=14.89, adjusted p=0), while their use of Strategy 6 (query preparatory) was significantly fewer than the expected value ( z=-3.95, adjusted p=0.004). Those with a score of 6, 4, and 3 had fewer NAs than the expected values ( z=-3.73, adjusted p=0.011 [Score 6]; z=-3.41, adjusted p=0.038 [Score 4]; z=-3.73, adjusted p=0.011 [Score 3]).

There was some variation in the use of common strategies depending on the question, but in general, participants with higher scores tended to use more modifications in addition to the main strategy. For Question 1, Strategy 6 (query preparatory) was the most commonly used main strategy overall. This was often combined with Strategy 4 (want statement) for participants scoring 6, Strategy 2 (performative) for participants scoring 3, and Strategy 1 (imperative) for participants scoring 2. Furthermore, the combination of modifications, Strategy 10 (promise) and Strategy 2 (reason), was common for participants with a score of 3 or higher. For Question 2, Strategy 6 (query preparatory) was the most popular among participants with a score of 5 or 6, while participants with a score of 6 combined this strategy with Strategy 5 (direct question) or Strategy 9 (mitigated expression). Participants with a score of 3 tended to use Strategy 1 (hedging), while participants with a score of 2 tended to use Strategy 5 (direct question). Higher-scoring individuals tended to combine a variety of modifications in addition to main strategies. Participants with a score of 5 used Strategy 2 (reason), Strategy 3 (apology), and Strategy 4 (preparator), while participants with a score of 6 combined the above strategies with Strategy 7 (minimizer) and Strategy 13 (attention getter). Participants with scores under 4 tended to use fewer modification strategies; however, participants with a score of 3 were found to prefer Strategy 13 (attention getter), while participants with a score of 2 were more likely to either use Strategy 4 (preparator) or no modification strategy at all. Finally, for Question 6, participants scoring 3 and 4 used both Strategy 4 (want statement) and Strategy 6 (query preparatory) for their main strategies. Participants who scored 5 varied too greatly for the measurement of any trends but participants who scored 6 used Strategy 6 (query preparatory). Participants scoring 3 or more used Strategy 2 (reason). Participants scoring 3 and 5 also used Strategy 3 (apology) as modifications, while participants scoring 6 used Strategy 4 (preparator) and Strategy 9 (question). Participants scoring 3 or 4 used Strategy 8 (external “please”), while participants scoring 3 had the tendency to not use any modifications as well. Individuals with lower scores tended to use fewer strategies. For Questions 1 and 2, participants with scores of 1 unanimously used no strategies at all, while participants who scored 2 or lower tended not to use any main strategies or modifications for Question 6.

Apology

Apology in Q5, Q7, and Q10 were coded from the main strategies only. Table 5 and Figure 4 show the relationship between pragmatic scores and the main strategies used in apology.

Figure 4 indicates that Strategy 1 (apology) and Strategy 3 (explanation) were extensively used, especially by those who scored 4. Although Fisher’s exact test scores confirmed that the score groups and strategies used were significantly different ( p=0), the post hoc analysis with Bonferroni’s adjustment showed that those who scored 5 used Strategy 4 (compensation) more than the expected value ( z=3.59, adjusted p=0.013). Although no other score groups revealed statistically more frequent use from the overall analysis, those who scored 6 used Strategy 4 (compensation) significantly more than the expected value ( z=3.57, adjusted p=0.001) in Q7 and Strategy 5 (promise for the future) ( z=3.31, adjusted p=0.033) in Q10. As with other speech acts, the number of NA was significantly higher for those who scored 1.

Refusals

Refusal strategies in Q3, Q8, and Q9 were coded from the main strategies and modifications. Table 6 and Figure 5 show the relationship between pragmatic scores and main strategies, and Table 7 and Figure 6 show the relationship between pragmatic scores and modifications used in refusals.

Fisher’s exact test results showed that the score groups and main strategies had a significant difference in refusals ( p=0.0005). As with other speech acts, the main strategies implemented varied according to the questions. The post hoc test with Bonferroni’s adjustment showed that those who scored 3 showed significantly greater usage of Strategy 4 (indefinite reply) than the expected value ( z=4.25, adjusted p=0.001) in Q3. In Q8, those who scored 6 frequently used Strategy 9 (Alternative), ( z=4.31, adjusted p=0). For all questions, those who scored 1 had an NA significantly greater than the expected value. Although Figure 3 indicates that Q3 saw Strategy 5 (excuse) used the most for respondents that scored 3 or higher, this was not supported by the post hoc analysis with Bonferroni’s adjustment. Likewise, Question 8 saw Strategy 2 (negative willingness/ability) as the most commonly used main strategy for responses that scored 2–5 while Strategy 1 (non-performative “no”) was used for respondents that scored 3, 4, and 6. Q9 saw Strategy 8 (avoidance) used most for scores of 3 and higher, while responses that scored 5 and 6 also used Strategy 9 (Alternative). However, these differences were not statistically significant.

Regarding modifications, the overall analysis showed that the score group and strategy used were significantly different ( p=0.0005). When we conducted the Fisher’s exact test for each question, those who scored 3 showed significantly greater use of Strategy 4 (apology) than the expected value (z=4.25, adjusted p=0.001) in Q3.

The common modifications to Q8 were Modification 4 (indefinite reply) and Modification 5 (reason) for scores of 3, 4, and 6. The most commonly used modification for Q9 was Modification 5 (reason) for responses that scored 3 or higher, while responses that scored 6 also used Modification 1 (positive comment). As with other speech acts, responses that scored 1 or 2 tended not to use any main strategies or modifications, which was supported statistically.

Advice giving

Advice giving in Q4, Q11, and Q12 was coded from the main strategies only. Table 8 and Figure 7 show the relationship between the pragmatic scores and strategies used in advice giving.

Fisher’s exact test confirmed that the score groups significantly differed in strategy use in advice giving ( p=0). The post hoc test with Bonferroni’s adjustment showed that those who scored 6 used Strategy 3 (indirect) ( z=3.30, adjusted p=0), those who scored 5 had Strategy 2 (softened) ( z=3.67, adjusted p=0.005), and those who scored 3 used Strategy 1 (direct) ( z=4.56, adjusted p=0) significantly more than the expected value, respectively. The frequent use of Strategy 1 by those who scored 3 was statistically significant only for Q11 ( z=3.15, adjusted p=0.039). As with other speech acts, those who scored 1 showed significantly more NAs, regardless of questions.

In this section, we focus on 30 participants who had worked overseas and analyze their survey responses by classifying the countries in which they worked into the Three Circle Model of World Englishes by Kachru (2019): (a) Inner Circle, such as America and the UK, (b) Outer Circle, such as Philippines and Singapore, and (c) Expanding Circle, such as China and (South) Korea.

The first survey question asked if they had ever faced difficulties communicating with colleagues who had cultural and linguistic differences. Of the 30 participants with experience working overseas, 21 (70%) responded “Yes,” while 9 (30%) responded “No.” They also gave examples of such challenges, and their open-ended responses were categorized as (a) word choice, (b) direct/indirect expressions, (c) understanding jokes and implicatures, (d) accent or intonation, and (e) others. (a) Lack of word knowledge was reported by both those who worked in Inner Circle countries and those who worked in Outer/Expanding Circle countries. Specifically, the participants felt difficulty if Japanese expressions did not have equivalent ones in English. Subtle differences in nuance were also highlighted by both participants who had worked in Inner Circle countries and those who had worked in Outer/Expanding Circle countries. Next, (b) direct/indirect expressions were mentioned by three participants who had worked in Inner Circle countries. Two participants experienced language breakdowns, which were caused by their periphrastic utterances. They represented politeness in Japanese but were not understood by their colleagues. Another participant unnecessarily downgraded himself when he asked his colleague to do something so that he would not make his colleague uncomfortable. As a result, his colleague wondered if the participant was bullied in the workplace. Two participants who worked in Outer/Expanding Circle countries also mentioned (b) direct/indirect expressions. One intentionally used direct expressions to avoid a misunderstanding, but she felt that she was being impolite in that moment. Another participant was puzzled when workers estimated their own skills/knowledge. She stated as follows:

Japanese people tend to underestimate their own skills, so they are often unassertive, even when we are confident with a certain task. In contrast, some people tend to overestimate their skills and say, ‘I can do it!’ with confidence. Such cultural and linguistic differences can cause difficulty in estimating who can do a certain task and to what extent.

(c) Understanding jokes and implicatures was also highlighted by two participants who had worked in both Inner and Outer/Expanding Circle countries. Although this is not the focus of this study, four of 21 participants who experienced difficulties listed (d) accentedness. (e) Others highlighted the timing of turn-taking and perceptions toward Japan and the Japanese people.

The second question asked what types of strategies the participants used to avoid miscommunications with colleagues from different backgrounds. This question had the following answer options, and the participants were able to choose all the options that applied to them. The results are shown in Table 9 and Figure 8.

The Fisher’s exact test results showed that there was no significant difference in frequency ( p=0.662). There is a possibility that increasing the number of participants who have worked overseas will reveal differences in intentional strategy use.

The final survey question asked what participants believed was necessary to work abroad. The responses did not differ regardless of whether they had worked in Inner Circle or Outer/Expanding Circle countries. Communicating with local associates, which does not necessarily mean language proficiency, was evaluated by ten participants, and understanding of religions and cultures was also mentioned by eight participants. People might assume that English proficiency is the key to success when working overseas, but less than half of the participants who had worked overseas (eight out of 30) mentioned English proficiency regardless of where they worked. Four participants claimed that fluent English is not required to work abroad. Eight participants believed that being flexible and tough is crucial, and two insisted that having special knowledge is more important than language skills because such specializations can mitigate language deficiencies.

The purpose of this study was to examine the pragmatic strategies produced by Japanese employees who use English at work. We focused on four speech acts: requests, apologies, refusals, and advice giving. The relationship between pragmatic scores and strategies was scrutinized quantitatively and qualitatively. Here, we summarize our findings with reference to the RQs.

Regarding RQ1 (What influences the pragmatic score more: frequency of strategy use or individual difference factors?), we found that those who obtained higher pragmatic scores tended to use more than one main strategy and multiple modifications regardless of the speech act type. In other words, speakers attempt to explain, make positive comments, use mild expressions, and produce a certain speech act clearly. The impact of the experience of working overseas was not as large as expected, but it affected the pragmatic scores of refusals. English proficiency partially affected the production of apologies. How often the participants encountered each situation was not a significant factor.

As for RQ2 (What types of pragmatic strategies lead to higher pragmatic scores?), the use of indirect strategies was key. In requests, many participants asked, “Could you … ?” (Query Preparatory). Whether they obtained a high score of 5 or 6 depended on whether they also had sufficient context for the request act, such as explaining the reason the speaker would like the addressee to do something and promising that it would lead to a positive result. Only those who marked the full score tended to use mitigated expressions ( e.g., “I'm wondering if …”). Direct strategies, such as the use of imperatives ( e.g., “Please answer …”), did not lead to higher scores, and many who used this strategy obtained a score of 3. In an apology, the use of an apology and reason was popular across the score groups. However, those who obtained a score of 5 or 6 tended to use remuneration in addition to apology-related expressions. Refusals were the most difficult speech act for the individuals in this study, and those who scored a score of 3 tended to use an apology ( e.g., “Sorry”) without clearly refusing. In contrast, those who obtained a higher score used an alternative, such as offering another day. Advice giving showed a clear distinction among the score groups. Those who garnered a score of 6 frequently used the Indirect strategy, those who scored 5 used the softened strategy, and those who scored 3 extensively used direct expressions.

Concerning RQ3 (What perceptions do Japanese workers who have been employed in foreign countries have toward the intercultural workplace communication?), we targeted 30 people with experience working overseas. There was no statistical disparity depending on where they worked, but qualitative analyses indicate that there were some distinctions regarding directness that were noticed between Japanese participants who worked abroad in Western contexts where English was likelier to be the sole language, unlike participants who worked in Asian contexts where multiple languages were likely to be utilized. Japanese participants who worked in Western contexts tended to use indirect strategies with longer responses and multiple strategies and modifications. Conversely, Japanese participants who worked in Asian contexts tended to utilize shorter responses with more direct approaches and fewer strategies.

Newton and Kusmierczyk (2011) mentioned, “Choosing an appropriate level of directness in status-differentiated interactions involving requests and refusals or participating in small talk requires sensitivity to cross-cultural differences in the way talk functions in different work settings” (p. 81). Building on this and other previous research, it could be argued that the level of directness used is a sociopragmatic skill that might have varying approaches depending on the cultural context in which English is employed. Thus, when considering these disparities, we can only consider that overly direct short responses with fewer strategies employed might not be sufficient in a native-speaker context.

Although we have discussed general tendencies, there were some variations within the speech acts. For example, in requests, Strategy 6 (query preparatory) was the most frequently used main strategy for subjects, with a score of 5 or 6 in Q1 and Q2. Nevertheless, this strategy was not used in Q6 compared to Q1 and Q2. In addition, Strategy 9 (mitigated expressions) was frequently used for those who scored 6. Participants used this main strategy in Q2, but this strategy was not so common in Q1 and Q6.

Subsequently, we considered the possible reasons why individual difference factors did not affect pragmatic scores compared with the frequency of strategy use. Regarding English proficiency, self-reported proficiency was not as reliable. It was impossible to include a proficiency test in addition to the 12-item DCT within the time limit, which we could post as a task on CrowdWorks. However, it would have been better if we could have included a single proficiency index that accurately measures participants' English proficiency. Regarding the experience of working overseas, the average score of those who had worked overseas was higher than that of those who had never worked overseas. Nonetheless, their work experiences might have correlated with their English proficiency in this study. Even if some companies do not regard workers’ English proficiency as the most significant factor when choosing who should be transferred abroad, it can be one of the indicators of a good candidate. Some participants might have studied English extensively when they were students so they could work overseas in the future. By including two groups with the same English proficiency test scores—one with experience working overseas and the other without such experience—it will be possible to determine if overseas work experience affects the production of speech acts.

As discussed in the methods section, interrater reliability could be achieved between the two raters. For most participants’ responses, the two raters agreed on the scores to be coded in terms of grammar and social nuance. However, a few differences were observed in scoring methods. The Japanese rater tended to rate some responses slightly more generously by understanding the cultural norms of Japanese workplaces. This is mainly because they could imagine a phrase as a direct translation from Japanese and regarded it as a common expression in Japanese offices, particularly among higher-ranking individuals. The American rater, though, felt that these expressions would not be as widely accepted beyond the scope of Japan, requiring too much understanding of Japanese culture to fit intercultural contexts abroad. Consequently, through discussion and comparisons of these responses, the two raters could agree on appropriate scores that reflected how expressions would be accepted in a wider intercultural context, with consideration to how expressions would be interpreted by a wide range of speakers with different backgrounds, including native English speakers.

The study had several limitations. First, we collected data through a written DCT, in which subjects wrote down the expressions in a specific task. We selected this method to reach out to many Japanese employees who have worked overseas or those who are currently working abroad. An online survey was necessary to reach out to working adults for a quantitative study. However, as many researchers have claimed ( e.g., Ishihara & Cohen, 2022; Taguchi, 2022), data from written DCTs do not necessarily reflect the actual language that participants use. Although spoken DCTs are more challenging for crowdsourcing studies, collecting spoken data is required for future studies. Second, individual difference factors must be controlled for more thoroughly. Some crowdsourcing services abroad specified for research participation ( e.g., Prolific, see Palan & Schitter, 2018) have many prescreening options, such as industry, position, and customer service experience. However, we could not filter potential participants from several perspectives in our study. In our research advertisement, we emphasized that we were recruiting people who used English at work and/or people who had experience working overseas. Furthermore, we added an option so 100 potential participants who had English skills would see our research advertisement. Nonetheless, those who were not confident in their English skills also joined our study, and only one-third of the participants had experience working overseas. To confirm the difference between those who have worked in inner circles and those who have worked in outer/expanding circles, more research is necessary to determine whether such responses would be acceptable in Asian contexts. Third, pragmatic factors, such as “power relationship (P),” “social distance (D),” and “degree of imposition (I)” ( e.g., Taguchi, 2022), were not thoroughly controlled in this study. It has been illustrated that PDI influences speech act strategies. However, due to the number of factors in this study, we could not include these factors in the research design. It would be desirable to control these levels and examine the relationship between PDI and industry.

Despite the limitations above, we revealed that pragmatic strategies lead to more acceptable speech. We propose the following pedagogical implications based on our findings. First, English teachers in Japan should explicitly explain to students that using direct speech is not always the best choice. Meaning should be conveyed clearly, but there are several ways to show politeness. For example, teachers can notify students that speech acts can be produced appropriately if each main strategy is accompanied by modifications. Successful communication should be addressed as a whole, not just as a sentence. Second, showing sample expressions that can be used in refusing someone’s offer in several contexts would help learners. This is because the comparison between speech act types depicted that requests were the easiest, while refusals were the most difficult. Third, in the EOP context, it would be important to teach not only English expressions but also cultural disparities. For example, if the learner were to work in a country where offering compensation in an apology is regarded as inappropriate, he/she should follow the custom. Additionally, one expression can be offensive if a subordinate produces it to a boss, even if the opposite is acceptable. Such knowledge is useful for Japanese employees who work abroad and for multicultural corporations.

We scrutinized the pragmatic strategies produced by Japanese employees. Some people in Japan might believe we must be direct when we communicate with people in other countries in English, but direct strategies were mostly found in those who scored 3 and 4, not those who scored 5 and 6. In other words, to mark the full score, speakers use indirect strategies, softened strategies, and a softening context. We hope that this study sheds light on EOP research for Japanese workers who conduct business in English.