2. The status of the native norm at the emergence of LCR

The end of the 1980s and the beginning of the 1990s witnessed a ‘paradigm shift’ (Kachru, Reference Kachru1991, p. 5), epitomized by the Quirk-Kachru controversy, which Kachru (Reference Kachru1991) described as an opposition between ‘deficit linguistics’ and ‘liberation linguistics’. Quirk (Reference Quirk1990) saw standard native English as the only acceptable variety. As a proponent of ‘deficit linguistics’, he considered that ‘[i]t is neither liberal nor liberating to permit learners to settle for lower standards than the best’ (Quirk, Reference Quirk1990, p. 9). Kachru (Reference Kachru, Quirk and Widdowson1985), on the other hand, as a representative of ‘liberation linguistics’, claimed that the English language is not owned by native speakers and that ‘English is now the language of those who use it’ (p. 20). In particular, he promoted institutionalized second-language varieties, those non-native varieties found in former British or American colonies where English has a (semi-)official status, and underlined their norm-developing nature. Around this period, several publications appeared that questioned the place of the native speaker: The native speaker is dead! (Paikeday, Reference Paikeday1985a); May I kill the native speaker? (Paikeday, Reference Paikeday1985b); Displacing the ‘native speaker’ (Rampton, Reference Rampton1990); ELT: The native speaker's burden? (Phillipson, Reference Phillipson1992); Going beyond the native speaker in language teaching (Cook, Reference Cook1999).

It is in the context of this ‘paradigm shift’ that LCR emerged in the late 1980s. With its focus on non-native varieties, LCR was well anchored in the new paradigmatic context, thus departing from the general field of corpus linguistics at the time, which ‘had shown great potential in investigating a wide range of native language varieties ( … ) but had neglected the non-native varieties’ (Granger, Gilquin, & Meunier, Reference Granger, Gilquin, Meunier, Granger, Gilquin and Meunier2015a, p. 1). In fact, the International Corpus of Learner English (ICLE), one of the first learner corpora, initially started as part of the International Corpus of English (ICE), a project that is essentially known for its contribution to the study of institutionalized second-language varieties and that is hence a prime example of ‘liberation linguistics’. Unlike the norm-developing institutionalized second-language varieties, however, the foreign-language varieties investigated in LCR are said to be norm-dependent (Kachru, Reference Kachru, Quirk and Widdowson1985), which necessarily brings the native norm more to the fore.

In addition to this special context, LCR, because of its focus of interest, also developed links with the field of second language acquisition (SLA) research, which at the time was influenced by two opposing forces. The first one was the (dominant) tradition of analyzing learner language relative to a native norm, as illustrated by the application of notions such as ‘error analysis’, which seeks to identify non-native features in learner language, or ‘fossilization’, which characterizes a stage in language development that fails to reach a native level. According to this tradition, summarized by Cook (Reference Cook1997, p. 38) in a literature review, ‘[t]he L2 learner's goal is ( … ) “full native speaker competence”, paraphrased as “perfect command” or “language mastery”. L2 learners are failures because they do not attain the same competence as native speakers’. The second force in SLA was the idea of comparative fallacy (Bley-Vroman, Reference Bley-Vroman1983), which advocates the study of learner language as a system with its ‘own internal logic’ (p. 14). Comparisons of learner language with native language are rejected on the grounds that ‘the learner's system is worthy of study in its own right, not just as a degenerate form of the target system’ (p. 4). Interestingly, these two approaches are reflected in one of the earliest and most popular methodologies in LCR, namely contrastive interlanguage analysis (CIA; Granger, Reference Granger, Aijmer, Altenberg and Johansson1996). This methodology consists of two types of comparison: (i) a comparison between native and non-native language, which corresponds to the traditional approach of judging learner language against a native norm, and (ii) a comparison between different types of non-native language, which does not involve any native reference, following the precept of comparative fallacy.

3. Does LCR require a native norm?

Despite a context that could have favoured both a ‘liberation’ approach and a ‘deficit’ approach, it must be recognized that the native norm has always occupied a major role in LCR. Flowerdew (Reference Flowerdew, Granger, Gilquin and Meunier2015, p. 469) points out that ‘[a] key facet of learner corpus research is that the learner corpus is usually compared with a native-speaker control corpus’. Typically, the results obtained through the analysis of learner corpus data are checked against the results from native corpus data representing the target language, which makes it possible to identify traces of non-nativeness, including errors, but also cases of under- or overuse, that is, the presence of fewer or more instances of a certain feature in learner language than in native language. To assess the role of the native norm in LCR, I have carried out a literature review on the basis of the 57 publications that are described as representative studies in the chapters of the Cambridge handbook of learner corpus research (Granger, Gilquin, & Meunier, Reference Granger, Gilquin and Meunier2015b). Of these studies, 32 (56%) involve a comparison of the learner data with a native norm, while the others do not involve any such comparison. However, of these remaining studies, 15 (26%) still refer to and somehow rely on a native norm, for example by identifying and sometimes annotating errors, or by using a native corpus as a resource from which word frequencies are extracted or collocations are checked. All in all, there are only ten studies out of 57 (18%) that do not include any kind of native reference.

The use of a native norm in learner corpus studies, however, does not mean that learner language must be seen as ‘deficient’ if it differs from native language, since the norm need not be prescriptive. Hunston (Reference Hunston2002, p. 212) observes that ‘information about differences between learner and native-speaker usage is not necessarily taken as a criticism of the learners involved’. The terms ‘underuse’ and ‘overuse’, for example, are merely descriptive concepts referring to a statistically significant difference in frequency between native and non-native language.

In addition, the norm used in LCR has the advantage of being abstracted from corpus data and hence ‘based on the performance of many individuals in various communication situations’ (Mukherjee, Reference Mukherjee2005, p. 14). As explained by Klippel and Mukherjee (Reference Klippel, Mukherjee, Volk-Birke and Lippert2007), this results in a norm that is empirical and frequency-based, and that therefore represents an objective yardstick to measure learners’ progression towards a certain target. The norm in LCR also tends to be explicit, being clearly identified as such in LCR studies. This is to be contrasted with some SLA studies that allegedly avoid the trap of comparative fallacy but actually rely on some hidden native norm, which leads Granger (Reference Granger and Aijmer2009, p. 18) to talk about ‘comparative hypocrisy’ in SLA.

The objectivity and explicitness of the native norm in LCR does not necessarily imply that such a norm is required or, indeed, desirable. The so-called ‘English Native Norm Irrelevance Hypothesis’ (House, Reference House, Locher and Strässler2008, p. 355) put forward by some advocates of English as a lingua franca (ELF) suggests that native models might not be relevant (see also Jenkins’ (Reference Jenkins2000) quote in Section 1). Amorim (Reference Amorim2013, p. 111), writing from an ELF perspective, claims for example that ‘[w]hen the aim is practising communication, the native speaker “yardstick” ( … ) seems inappropriate’, adding that ‘EFL learners do not aim at becoming mother tongue imitators’. This may be true of certain populations of non-native speakers, and in particular L2 users rather than learners (to use Mauranen's (Reference Mauranen, Meunier, De Cock, Gilquin and Paquot2011) distinction), but for many learners, the preferred norm is still native. Every year at the beginning of a course I teach on English varieties, I ask my students what norm they would like to approximate to. They always predominantly choose a native variety. Between 2008 and 2020, 55% (out of a total of 406 students) chose British English and almost 20% chose American English, plus some 2% for Canadian English and 2% for Australian English. About 12% chose ‘English as it is used by competent non-native speakers in international contexts’Footnote ² as their preferred target. This tendency was even more striking when the students were asked what English norm they would teach. Some 67% chose British English and 24% chose American English, leaving less than 5% for competent non-native use. This strong preference for one of the two main native varieties is probably partly due to the fact that most of these students are English majors, but since many of them are to become English teachers, their preferences are important as they might influence the variety of English that future students will learn.

While we have seen that a (corpus-based and explicit) native norm is the rule in LCR and that it makes sense in view of (certain) learners’ needs and wishes, it must be underlined that, as already suggested, LCR studies can and do dispense with a native norm. In the CIA model described earlier, the comparison between learner and native corpus data is only one of two components. The other one is a comparison of different types of learner corpus data, such as data produced by learners with different L1s (the most frequent kind of comparison for this component), data produced by learners with low vs high exposure to the target language, data produced by learners at different stages of acquisition, data produced in timed vs untimed conditions, etc. This makes it possible to study learner language independently of any native norm. Vyatkina (Reference Vyatkina2013a), for example, adopts a longitudinal design to compare the written production of learners over four semesters and 14 measurement occasions. This allows her to provide a fine-grained analysis of the development of syntactic complexity among the learners, without comparing them against any native baseline; the point of reference is that of the previous measurement occasion(s). LCR can thus do without a native norm and investigate learner language in its own right, as advocated by certain SLA researchers.

4. What corpus-derived norms are available and how do we choose?

One of the lessons of corpus linguistics is that a language, especially an international language like English, is far from being a monolithic entity. Even if we limit ourselves to native English, there is a range of varieties, and hence a range of norms, to choose from. These norms vary according to factors such as geography, social class, age, medium or formality. I will deal with three factors here: regional variety, level of literacy and text type. I will then review some potential non-native norms and will propose general criteria to choose the most appropriate norm(s).

5. What do we do with these norms in the classroom?

From the early days of LCR, its potential for language pedagogy has been emphasized. While concrete pedagogical applications of LCR so far have not been as numerous as we might have hoped, its pedagogical implications are widely recognized and have created strong links between LCR and the field of FLT (see, e.g., Meunier, Reference Meunier, Farr and Murray2016). It therefore seems relevant, for the last question, to turn to the pedagogical implications of corpus-derived norms.

Let us first ponder whether the trend of liberation linguistics initiated by Kachru (see Section 2) should lead to a kind of ‘liberation pedagogy’. Not following any kind of norm does not seem like an option in the foreign language classroom, since a norm is required to decide what kind of language should be taught and to assess students’ progress towards the set target. However, we may wonder whether a linguistic feature in learner language can be considered acceptable as long as it respects some norm. Corpora give access to many linguistic norms, some of them characterized by features that are not normally found in standard native varieties, which would potentially make many learner features acceptable. Actually, corpus studies comparing learner English and New Englishes have brought to light non-standard features shared by the two types of varieties. In such cases, it may seem unfair to condemn learners who produce features accepted in certain varieties. For example, the use of discuss about has been described as an ‘interesting phenomenon at the lexis-grammar interface’ in Indian English (Mukherjee, Reference Mukherjee2007, p. 175) but is seen as an error in pedagogical materials for learners (e.g. Turton & Heaton, Reference Turton and Heaton1996). The main reason why such a difference in treatment may seem unfair, however, is that we look at it from our own perspective as linguists. Chances are that it does not seem unfair to teachers and learners, most of whom may not even be aware that discuss about is an innovative feature in certain New Englishes. Besides, learners who produce this phrase most likely do not produce it because they know that it is acceptable in Indian English, but because they think that it is correct in standard native English, probably by analogy with the noun discussion or with verbs like talk or speak (see Gilquin, Reference Gilquin, Evers-Vermeul and Tribushinina2017). This underlines the importance of making a distinction between research norms, those that linguists use, and so-called pedagogical norms, those that ‘guide the selection and sequencing of target language features for language teaching and learning’ (Bardovi-Harlig & Gass, Reference Bardovi-Harlig, Gass, Gass, Bardovi-Harlig, Pierce and Walz2002, p. 1). While the pluricentric, multi-norm model advocated by Kachru (Reference Kachru1991, p. 4) is appealing from a research point of view, it is ‘simply not a practical model for classroom teaching purposes’ (Ur, Reference Ur2010, p. 89). Teaching several norms in the classroom would involve a number of practical problems, including the fact that time constraints would not allow teachers to teach and learners to learn multiple norms thoroughly, that textbooks and other pedagogical materials almost exclusively conform to standard British or American English norms, and that the repertoire of norms that teachers master is necessarily limited. It is symptomatic, in this respect, that scholars who suggest using norms such as New Englishes or ELF in the EFL classroom usually do not offer much in the way of concrete proposals on how to do this (possible exceptions are Jenkins’ (Reference Jenkins2000) Lingua Franca Core, which is not without controversy though – see Ugarte Olea (Reference Ugarte Olea2019) – as well as the Global Englishes Language Teaching (GELT) framework and some other proposals summarized in Rose, McKinley, and Galloway (Reference Rose, McKinley and Galloway2020), which however seek to adopt a New Englishes or ELF approach rather than teaching a New Englishes or ELF variety per se).

In the same way as native norms are still relevant in LCR, they arguably still have a key role to play in the foreign language classroom, not only because many learners have native language as their target, as we saw earlier, but also because, in real life, learners are likely to be judged according to native standards. Think, for example, of standardized tests like TOEFL, which relies on a native norm. Even IELTS, which aims to treat ‘all test takers with the utmost fairness and respect by actively avoiding cultural bias’, accepts ‘all standard varieties of native-speaker English, including North American, British, Australian and New Zealand English’ (emphasis added).Footnote ³ The use of non-standard/non-native language features has also been shown to have negative consequences in job applications (e.g. McMasters, Reference McMasters2004) or publishing (e.g. Flowerdew, Reference Flowerdew2008). Teachers would therefore be doing learners a disservice by denying them access to native norms. Actually, a survey by Young and Walsh (Reference Young and Walsh2010) shows that the teachers who were polled almost exclusively taught American or British English and that none of them taught English as a lingua franca or as an international language.

This does not mean that we can simply take a reference corpus of native language and start teaching learners the usage it reflects. Prodromou (Reference Prodromou1998, p. 267) asks ‘how much “native-speaker reality” can the non-native speaker and the EFL classroom take?’. As I see it, there are at least two aspects to consider. The first one is that, as Widdowson (Reference Widdowson2000) emphasizes, native corpora may lack authenticity for learners, who may not relate to texts produced in a different, foreign context (although the compilation of reference corpora with data produced by subjects whose profiles closely match those of the learners may reduce this risk). The second aspect is that corpus-derived findings may be too detailed for learners. The information contained in the corpus-based Collins COBUILD Grammar Patterns volumes (Francis, Hunston, & Manning, Reference Francis, Hunston and Manning1996, Reference Francis, Hunston and Manning1998), for example, is extremely valuable from a research point of view, but the degree of detail is such that much of it would probably be lost on learners.

Another key issue about corpus-derived information is that it should not be acted on blindly. For one thing, native corpora, even expert ones, may include non-standard or non-politically correct features that teachers would not want their students to reproduce. For another, as noted by Granger (Reference Granger and Aijmer2009, p. 22), learner features uncovered by LCR ‘need not necessarily lead to targeted action in the classroom’. Deviations from the native norm, in the form of misuse, underuse or overuse, should be considered in relation to criteria such as learners’ needs, teaching objectives or teachability, and on that basis should be selected or disregarded for classroom action (p. 22).

It was suggested earlier with respect to research that the norm should ideally be adapted to learners’ needs. This is true of teaching too. Corder (Reference Corder1981, p. 49) already said so when he recommended ‘adopting more realistic norms/standards given the particular sort of student we have, or promoting alternative norms/standards for some subgroup of the student population’. However, since mixed classrooms have become the rule rather than the exception, it is almost impossible for teachers to use tailor-made norms, since what one student wants or needs is not necessarily what the other students want or need. A one-size-fits-all norm may therefore be the only practical solution. In that case, choosing among the major native varieties (mainly standard British or American English) arguably makes it possible to opt for a norm that is likely to be most useful to a majority of the learners and/or most familiar to the teacher. This norm can then be imposed more or less strictly depending on the context. For example, it may be less strictly imposed on students learning English for Occupational Purposes (e.g. business English) than on students specializing in the study of English. It may also be less strictly imposed in oral communication, where communicative competence and mutual intelligibility may prevail, than in written usage, where linguistic competence and language accuracy may be more crucial.

I would like to stress that this choice of a single native norm does not imply that other norms should disappear from the classroom altogether. If possible at all, learners should be made aware of the existence of some of these other norms and they should be taught to respect these norms, and possibly understand them, for example through the use of audio and video materials in different language varieties. This, incidentally, points to yet another relevant distinction, namely between productive and receptive norms.

Finally, it is important to underline that the use of corpus-derived norms should not deny learners their right to creativity. Kreyer (Reference Kreyer2015, p. 161) asks whether we should view the norm as ‘binding’ or allow for some flexibility, letting learners ‘experiment with the L2 and, as a result, come up with something that in part is different from the target norm’. That learners are capable of creative language use, just like native speakers, can be illustrated by the following example, taken from Gilquin (Reference Gilquin2015, p. 107): … because of the instructions to fashion your jeans up by stone-washing … (ICLE-GE). The phrasal verb fashion up, which one is unlikely to find in a native reference corpus (and which is not listed in any dictionary of the English language), is nevertheless perfectly intelligible. This simple example shows that ‘linguistic creativity is not the sole preserve of (norm-providing) native speakers’ (p. 107), an essential point to bear in mind when evaluating learners’ production.