Japanese EFL undergraduate students’ use of the epistemic modal verbs may, might, and could in academic writing

1 Introduction

Modal meanings include “epistemic possibility”, which we define here following Downing (2015: 343) as those that “assess the possibility, probability or otherwise of a state of affairs, according to the speaker’s limited knowledge or belief”. Numerous grammarians argue that modal verbs for epistemic modal concepts serve an important function in academic writing to avoid the impression of too much certainty and soften assertions (e.g., Becker and Feng 2020; Depraetere and Langford 2020; Hinkel 2020; Leech et al. 2009; Poole et al. 2019). Many grammarians also agree that modal verbs are difficult for ESL/EFL students to master due to their polysemy and semantic overlap (e.g., Larsen-Freeman and Celce-Murcia 2016).

General observations about the importance of epistemic modality in academic writing, difficulties this area may present for learners, and the lack of research specifically on the Japanese context motivated this corpus-based study on the three main modal verbs which are used to express “epistemic possibility”: may, might, and could.^[1]

There is a substantial amount of scholarship on the nature of academic writing more generally by English learners with different L1 backgrounds that points to learners using language that is more reminiscent of informal and/or spoken language (e.g., Aijmer 2002; Altenberg 1997; Breeze 2008; Gilquin and Granger 2015; Gilquin and Paquot 2008; Granger and Rayson 1998; Leedham 2011; McCrostie 2008; Özhan 2012; Petch-Tyson 1998; Tåqvist 2018; Virtanen 1998). Our hypothesis, based on the preponderance of studies that suggest that learners tend toward insufficiently formal use in their academic writing, is that this may apply to Japanese students’ use of modals as well. Most learner corpus studies have analyzed advanced learner data but Gilquin and Granger (2015: 419) have encouraged looking at learners at other levels. Compared to the CEFR levels of learners in many European countries, Japanese learners’ levels tend to be significantly lower, with the majority at A2 and B1 (Ishikawa 2015: 104; Tono 2020: 25). Regarding learners’ writing as a target of corpus research, scholars usually analyze discipline-specific writing rather than general essays or reflective writing (Aull 2019: 143), and little attention has been paid to registers other than reviews, theses and dissertations (Becker and Feng 2020: 257).

This study analyzes paragraph writing assignments in academic writing courses using a corpus compiled of data from Japanese undergraduates who fall roughly into the CEFR A2 and B1 levels. By focusing on the use of some modal verbs in academic paragraph writing (which has been studied less than other text types) by lower-level learners (who are still under-researched), we aim to contribute to understanding how these learners use modal verbs and what instruction may be helpful to them. This study will also use an American English (AmE) academic prose corpus. As we discuss in Section 3.1, we do not assume that the students see professional writers’ prose as their target. However, this corpus will be useful to gain insight into authentic usage of the three modal verbs. The reason for choosing AmE as our reference point is that English teaching in Japan is mostly oriented toward this variety. Consider for instance that in the Japan Exchange and Teaching Programme, 55.22 % of Assistant Language Teachers in 2022 were US nationals (JET 2022). Most English textbooks and English-Japanese dictionaries in Japan also privilege AmE, for example, with regard to spelling and pronunciation.

Based on our corpus findings and considering the risk of negative perceptions of English learners’ inability to use modal verbs appropriately, we argue that certainly in the Japanese context more attention needs to be paid to classroom instruction on these modals, especially could. We will discuss possible reasons as to why Japanese learners might deviate from AmE professional writers’ usage of may, might, and could in the ways they do, and suggest possible pedagogical implications for EAP instruction and teaching materials about these modals.

2 Previous studies of learners’ use of modal verbs

Aijmer (2002) compared Swedish, French, and German learners’ use of English in the International Corpus of Learner English (ICLE), which consists of argumentative essays by advanced proficiency level undergraduate students, with that of British university native speaker students in the Louvain Corpus of Native English Essays (LOCNESS).^[2] She reports French learners highly significantly overused may compared to native speaker students, but German and Swedish students did not show that pattern. This may suggest that the use of modal verbs varies depending on the learners’ first language. Aijmer also reports significant overuse of might across all three groups. As one of the possible reasons for this she mentions the influence of speech: might is more typical of conversation than of academic writing (cf. Holmes 1988: 29).

Aull and Lancaster (2014) analyzed first-year university students’ argumentative essays in a large public research university and a smaller, private liberal arts university in the US, and found that in the essays from the latter university modal verbs were used less frequently as hedges than approximative hedges including adjectives and adverbs (e.g., apparent, generally). They suggest that in order to expand the repertoire of stance features, such as hedges, students should be exposed to the characteristics of university and/or disciplinary writing (2014: 160).

Becker and Feng (2020) analyzed the use of modal verbs in the Physical Sciences subcorpus of the Michigan Corpus of Upper-Level Student Papers (MICUSP).^[3] This subcorpus was compiled from A-graded course assignment papers from final-year undergraduates through third-year graduates. Unlike graduate students, undergraduates used fewer possibility modals (including may, might, and could), which soften assertions, than prediction modals (e.g., will, would), which work in reverse. In this regard, Becker and Feng’s (2020) results confirm the findings of previous studies, that it is difficult for learners to adequately express the degree of certainty and that learners, without sufficient academic writing experience, tend to use boosters rather than hedges (e.g., Aull 2015, 2019; Hyland 2016).

Thus, there are studies analyzing the use of modal verbs at different learning stages. However, there is little research on the use of English by lower-level learners at the A2 or B1 CEFR levels, and still fewer studies have comprehensively discussed the factors that contribute to learners’ use of modal verbs.

Our research questions are as follows:

1. Are there any similarities and differences in the use of the modal verbs may, might, and could between Japanese learners of English and AmE expert writers?

2. If there are any differences, how might we explain them, and what might the pedagogical implications be?

3 Methodology

4 Results

Figure 1 shows the results of the corpus search of the three modal verbs with the “epistemic possibility” meaning, with normalized frequency (per 10,000 words). The difference in frequency of the modal verbs between the corpora was examined by log-likelihood tests (Rayson and Garside 2000). May with the “epistemic possibility” meaning seems to be used less frequently in JUEC than in AmE06_L, though the difference between the corpora is not significant. Might with this meaning appears more frequently in JUEC than in AmE06_L, but as is the case with may there is no significant difference. In JUEC, may is used with the meaning of “epistemic possibility” for 93.8 % (45 of 48) of all occurrences, and might is used with this meaning in 100.0 % of cases (26 of 26). The frequency of could is noteworthy. Could with the “epistemic possibility” meaning is used significantly less in JUEC than in AmE06_L (see Appendix B). Though the total frequency of could is much higher than that of may and might in JUEC, it is used with the “epistemic possibility” meaning in only 2.0 % of cases (3 of 148). This means that the Japanese students’ use of could merits careful attention, as it may be distinctive in some way(s).

Figure 1:

Frequency of may, might, and could with the “epistemic possibility” meaning in JUEC and AmE06_L.

The modal verbs in (3)–(8) are used with the “epistemic possibility” meaning in the two corpora:

Semantic analysis, based on Declerck (1991: 389–390), of Japanese students’ use of could shows that the “ability” meaning^[11] is significantly higher than in AE06_L. 89.19 % of all occurrences of could (132 out of 148 occurrences) in JUEC carry the “past ability” meaning, as in (9), with 11 tokens describing non-past ability (10):

Let us now look at the Japanese students’ use of could with the “ability in the past” meaning according to their English proficiency levels. One might hypothesize that as students progress in their overall proficiency, their use of could should gradually approximate that of the AmE writers. However, as Figure 2 shows, the Japanese students’ usage in this respect actually seems to move away from AmE writers’ usage.

Figure 2:

Could with the “ability in the past” meaning in JUEC and AmE06_L (frequency per 10,000 words).

Biber et al.’s (2021: 487) corpus search findings show that could is much more frequent in conversation than in academic prose, which in this grammar includes both British and American data. In order to focus specifically on AmE we carried out a search of COCA^[12] and found that in these data the frequency of could is much higher in spoken than in academic texts, the difference being 1429.44 versus 935.32 per million words. The Japanese students’ highly frequent use of could therefore seems an instance of informal usage.

We will consider the learners’ use of the “epistemic possibility” meaning of the three modal verbs in more detail. It is worth looking at how the learners progress with respect to their usage of these verbs in this manner. This has been visualized in Figure 3. Once again, one might have expected that Japanese learners, as they progress in their studies, would gradually approximate AmE writers’ usage. However, that does not appear to be the case: it is not really possible to detect any trend toward the AmE reference corpus. We conclude from this that the current approach to teaching this important use of these modal verbs does not seem to lead to increased approximation of AmE writers’ usage. (We have acknowledged, above, that perfect approximation is or should not necessarily be the learners’ goal, but it is an interesting observation nonetheless.)

Figure 3:

May, might, and could with the “epistemic possibility” meaning in JUEC according to English proficiency levels (frequency per 1,000 words).

5 Discussion and pedagogical implications

On the basis of our corpus search findings, the frequencies of may and might with the “epistemic possibility” meaning are not significantly different between JUEC and AmE06_L. The overall frequency of these modals in JUEC may provide some reassurance that these epistemic markers are broadly used appropriately, yet the lack of a clear direction toward AmE writers’ usage (see Figure 3) nevertheless raises questions about the way in which these verbs are taught. We will return to this later in this section.

The main observation to make, based on the data, concerns could. The Japanese learners use this modal significantly less often as an epistemic marker and very highly significantly more often to describe “past ability”, as in example (9). Whilst this example is perfectly acceptable, JUEC also contains instances of a rather different kind of “past ability”:

As Declerck (1991: 394) says, could “cannot be used to ASSERT the PERFORMANCE of a DYNAMIC situation on a SPECIFIC OCCASION in the past”. The difference between (9), on the one hand, and (11)–(12), on the other, relates to event structure: (9) describes an ability on the part of the subject (she), whereas (11)–(12) portray a specific event that the subject was not only able to do, but did in fact complete on one particular occasion. This use is not acceptable for could in American (or British) English, although it is for was/were able to and for the negated form could not. JUEC contains as many as 22 tokens like (11)–(12), which is a substantial subset of the total number of examples: 16.67 % of all past ability uses of could.

Having highlighted the main difference between JUEC and AmE06_L, we now consider what may underlie the Japanese learners’ deviations from the AmE writers’ usage of the modals, especially could. Gilquin and Paquot (2008), one of many studies to have identified the tendency for foreign learners to sound colloquial when attempting to produce English academic prose (see Section 1 for additional references), offer a useful overview of four factors that may explain this pattern: the influence of spoken language, L1 transfer, teaching-induced factors, and general lack of familiarity with academic writing.

Regarding the first factor, i.e., the influence of spoken language, Gilquin and Paquot (2008: 52) argue that in the context of EFL learning authentic spoken input, be it from teachers, teaching materials or exposure to English television and other media, mostly tends to be limited, and is therefore unlikely to play a very substantial role in explaining colloquial features in academic writing. They suggest that authentic input via mass media may be high in countries such as the Netherlands, but in Japan it is probably not the case.

L1 transfer, Gilquin and Paquot’s second factor, is probably significantly more important in the present study, especially in causing the learners’ use of (ungrammatical) “past performance on one occasion” could (see (11)–(12)). Japanese dekita can have the “past ability” meaning, like could, but it can also describe “past performance on one occasion”, unlike could; see the following examples from Nishitani and Nakazaki (2015; translation ours):

The seemingly significant impact of L1 transfer leads us to an important pedagogical implication of our study. Modal expressions, including modal verbs, display significant crosslinguisic variation in terms of their meanings and uses (see for example Nuyts and van der Auwera 2016). As EAP textbook authors would struggle to cover this variation we would suggest that there is an onus on instructors to identify the main differences between the L1 and English in this regard. This should be done using descriptive grammars of the languages and/or studies specifically on modal verbs. In view of the well-known importance of “noticing” in second language learning (cf. Schmidt 1990, 2012) instructors should draw students’ attention to these differences, and continue to be mindful of and highlight them when marking and providing feedback on students’ work.

As for Gilquin and Paquot’s (2008) third factor, teaching-induced factors, the authors’ examples of this are all inadvertent “distort[ions of] learners’ sense of what is stylistically appropriate or not” 2008: 55). This factor clearly contributes to Japanese learners’ use of could. An interesting question in relation to this is: is there anything in the teaching method that causes the Japanese students’ usage of this modal? This is not an easy question to answer comprehensively as it would require a detailed study of the teaching and learning process in its entirety, including interactions between teachers and students in class, written feedback, and so on. However, for the present purposes, let us focus on printed teaching materials which the Japanese learners in this study used: Zemach and Islam (2005, 2011; for details see Section 3.1) and in particular its coverage of could, which shows the most substantial deviations in the Japanese learner data from the AmE writers’ usage. The semantic analysis of all main clause instances of could in Zemach and Islam (2005), which was used by 55.7 % of the students (with the rest using a newer version of the same textbook), yields the following absolute and relative distribution of uses, with JUEC and AmE06_L included for the sake of comparison in Table 2 (normalized frequencies per 10,000 words not given because an EAP textbook, including instructions and in this case also word lists, is a very different genre than actual English academic prose).^[13]

The semantic analysis of Zemach and Islam (2005) tells us that insofar as this textbook is a reasonable proxy for the teaching as a whole, it can be held partly responsible for the deviations observed in JUEC from AmE usage, represented by AmE06_L. The textbook underrepresents the “epistemic” use and overrepresents the “ability” use, compared to AmE academic prose, which may help explain the Japanese learners’ linguistic behavior, although the learners appear to deviate from AmE for both meanings to a greater extent than the textbook. One pedagogical implication here may be that texts used in teaching materials would ideally reflect authentic usage to a higher degree.

Regarding “ability”, it is unsurprising but worth pointing out that the “ability” uses in Zemach and Islam (2005) do not contain any tokens of the unacceptable “past performance on one occasion” use (see (11)–(12) and associated discussion); we explain, above, that instances of this unacceptable use may be due to L1 influence and how this can be mitigated in teaching.

In addition to the actual language used by Zemach and Islam (2005), we may consider some grammatical topics they single out for special attention. For example, the authors offer quite an elaborate discussion of how to describe cause and effect (see their Units 7 and 11, which cover expressions such as because, so, therefore, and consequently). By contrast, there is no discussion on hedging one’s claims, which would have provided an appropriate context in which to explain the role of modal verbs and (some of) their (main) uses, especially in academic discourse.

Zemach and Islam’s (2005, 2011 disregard of modals and their use in academic prose is far from unique. Some examples of other textbooks from major publishers (all aimed at similar CEFR levels around A2 or B1) that display a similar gap are Blanchard and Root (2017), Folse et al. (2020), and Singleton (2022). However, one exception in this respect is Frodesen and Wald (2016). Each chapter in this book starts with a “raising language awareness” section, in which the authors draw attention to some lexicogrammatical structure and its importance in (academic) writing. Thus, Frodesen and Wald’s chapter that includes modals couches the discussion and exercises in a broader discussion of hedges (2016: 175–180). The authors observe that “[i]n academic writing, writers often need to qualify statements with vocabulary that expresses degrees of certainty or accuracy about the information they are conveying” (Frodesen and Wald 2016: 175). They single out modal verbs for special attention by discussing their use in academic prose in some detail (Frodesen and Wald 2016: 178–179) but their exercises also include epistemic stance adverbials such as typically and the copular verb seem (ibid.: 177), which commonly marks likelihood in academic writing (Biber et al. 2021: 438).

We would highlight Frodesen and Wald (2016: 175–180) as an instance of good pedagogical practice and would recommend that other textbook authors follow their example in relation to English modal verbs. Furthermore, our suggestion for EAP instructors is that they may wish to refer to Frodesen and Wald or prepare similar materials themselves.

Related to the lack of discussion of hedging, although Zemach and Islam (2005) clearly set out to develop students’ academic writing,^[14] the texts that students are asked to engage with and to produce include topics such as “gifts”, “a trend”, and so on. As these topics are not academic the texts are typically narrative – and it is well known that narrative texts differ in many respects, including the use of modal verbs, from academic prose (cf. Biber 1988).

Japanese students’ use of could, meaning “past ability”, in 11 different topics shows that the most frequent use is seen in their writing about “interesting or unusual experiences” (41 tokens, 31.1 %), followed by writing about “a difficult decision” (27, 20.5 %), and then in writing about “explanations and excuses” (19, 14.4 %). As for the topic with the highest use of could, the high frequency may be largely due to the purpose of writing a narrative story and the fact that students were likely to write about their personal experiences in the past. Even after excluding student writing on these three topics, the Japanese learners’ use of this modal is still significantly more frequent than that observed in the AmE data. Topics, then, may have an influence on Japanese students’ use of could. We would encourage EAP materials developers and instructors to focus more on academic topics and texts (e.g., on the relation between nutrition and health, or on climate change; cf. Frodesen and Wald 2016: 176, 179).

In explaining why EFL learners’ academic English is relatively informal, Gilquin and Paquot (2008) have made a useful start in unpacking teaching-induced factors into different types, specifically, (i) lists of expressions that appear to suggest full synonymy when there are in fact subtle semantic and/or register differences and (ii) translations of examples from the L1 into the L2 that are not sufficiently sensitive to register appropriacy in the L2 (2008: 54–55). To this we add three more types: (iii) significant lack of faithfulness to the register of academic discourse with respect to the frequency of a particular expression or usage in texts and exercises,^[15] (iv) lack of explicit discussion of an area of grammar that is important to academic prose and/or known to be difficult, and (v) lack of representation of academic texts.

The fourth and final factor that Gilquin and Paquot (2008) suggest may contribute to academic English produced by learners, is development, by which they mean development of the mastery of academic prose in learners’ L1. Based on data from the Louvain Corpus of Native English Essays (LOCNESS)^[16] they show that native speaker students often approximate expert academic writing in their L1 more closely than non-native students in their L2, but that there is nevertheless a noticeable tendency toward “spoken-like features” even in the L1 (2008: 56).

Studies on the acquisition of Japanese academic prose by Japanese university students (e.g., Akiyama 2021; Ishiguro 2011; Yamaji et al. 2013) suggest that there is a similar tendency toward speech-like features. For example, Ishiguro (2011: 15) notes that this tendency of the students seems to be accelerating with the spread of media such as text messages, blog, and Twitter in which spoken language is used. Based on these studies and our own, we suggest that when teaching Japanese university students, more emphasis should be placed on appropriate use of formal grammatical features, including modal verbs, in their academic prose both in Japanese and in English.

6 Conclusions

The analysis of our JUEC data suggests that the writing of the Japanese undergraduate students we targeted could be characterized as colloquial and not sufficiently academic, especially judging from the learners’ use of could. For may and might with the “epistemic possibility” meaning, we do not in fact find any significant difference between our learner data (JUEC) and the AmE reference corpus (AmE06_L). The Japanese students used could much more frequently than the AmE writers. However, they used it less frequently as a marker of “epistemic possibility”, which is an important use of this modal verb in academic writing. They used could very frequently with the meaning of “past ability”. This modal turns out to be especially problematic for Japanese learners.

To explain our findings, we turn to a set of four factors usefully summarized in Gilquin and Paquot’s (2008) study on the typically relatively colloquial nature of EFL learners’ academic prose. We conclude that the topics Japanese students are engaged with may cause (at least in part) their frequent use of could. L1 interference also seems to play an important role. Japanese dekita and could differ in the scope of their meaning, with dekita covering both “past ability”, which could can also signal, and the “past performance on one occasion” meaning, which in American (and British) English it cannot.^[17]

We have proposed a range of pedagogical implications. EAP materials writers and/or instructors should focus on differences between modals in the L1 and in English; explicitly cover modals as hedges and discuss hedging in academic writing; ensure that usage of modal verbs in teaching materials is analogous to authentic texts; select texts and topics that are academic in nature; and practice formal writing also in the students’ L1.