Corpus Research and Selecting What to Teach

Corpora have proven to be an invaluable resource in the design of language teaching syllabi that emphasize communicative competence (see Hymes, Reference Hymes, Pride and Holmes1972, Reference Hymes and Pütz1992) and give prominence to those items that occur frequently in the language and that learners are hence most likely to encounter in real-life communicative situations. In the context of computer corpus-informed ELT syllabi, the first and probably most groundbreaking development was the design of the Collins COBUILD English Course (CCEC) (Willis & Willis, Reference Willis and Willis1989), an offshoot of the pioneering COBUILD project in pedagogically oriented lexicography (see Sinclair, Reference Sinclair1987). The contents of this new, corpus-driven “lexical syllabus” are “the commonest words and phrases in English and their meanings” (Willis, Reference Willis1990, p. 124). With its focus on lexis and lexical patterns, the CCEC responds to some of the most central findings of corpus research, namely, that language is highly patterned in that it consists of an immense degree of repeated word combinations, and that lexis and grammar are inseparably linked. Also worth mentioning is a much earlier attempt to improve the teaching of English vocabulary that was made long before the advent of computers and electronic corpora. In 1934, Michael West organized a conference “to discuss the part played by corpus-based word lists in the teaching of English as a foreign language” (Kennedy, Reference Kennedy and Svartvik1992, p. 327). About 20 years later, West's (Reference West1953) General Service List of English Words (GSL) was published and has since then exerted great influence on curriculum design (cf. Kennedy, Reference Kennedy and Svartvik1992; Willis, Reference Willis1990). As the title indicates, West's GSL suggests a syllabus that is based on words rather than on grammatical structures. It is also based on frequently occurring words rather than on rare ones. Of course, frequency of occurrence is not the only criterion that should influence decisions about the inclusion of items in the teaching syllabus; there are other relevant criteria, such as “range, availability, coverage, and learnability (Mackey, 1965, p. 188)” (Kennedy, Reference Kennedy and Svartvik1992, p. 340; see also Nation, Reference Nation1990), but it is certainly an immensely important one (see also Leech, Reference Leech, Wichmann, Fligelstone, McEnery and Knowles1997). It can be safely assumed that learners will find it easier to develop both their receptive and productive skills when confronted with the most common lexical items of a language (and the patterns and meanings with which these items typically occur) than when the language-teaching input they get gives high priority to infrequent words and structures that the learners rarely encounter in real-life situations.

Another current strand in applied corpus research, which aims to inform the teaching syllabus and stresses the importance of frequency of occurrence, examines language items in actual language use and compares the distributions and patterns found in general reference corpora (of speech and/or writing) with the presentations of the same items in teaching materials (course books, grammars, and usage handbooks). The starting point for these kinds of studies usually includes language features that are known to cause perpetual problems for learners—for example, for German, discourse particles (Jones, Reference Jones, Wichmann, Fligelstone, McEnery and Knowles1997), modal verbs (Jones, Reference Jones, Lewandowska-Tomaszczyk and Melia2000), the passive voice (Jones, Reference Jones, Lewandowska-Tomaszczyk and Melia2000), and prepositions (Jones, Reference Jones, Wichmann, Fligelstone, McEnery and Knowles1997); and, for English, future time expressions (Mindt, Reference Mindt1987, Reference Mindt, Wichmann, Fligelstone, McEnery and Knowles1997), if-clauses (Römer, Reference Römer, Aston, Bernardini and Stewart2004a), irregular verbs (Grabowski & Mindt, Reference Mindt1995), linking adverbials (Conrad, Reference Conrad and Sinclair2004), modal verbs (Mindt, Reference Mindt1995; Römer, Reference Römer and Sinclair2004b), the present perfect (Lorenz, Reference Lorenz, Kettemann and Marko2002; Schlüter, Reference Schlüter2002), progressive verb forms (Römer, Reference Römer2005, Reference Römer, Teubert and Krishnamurthy2007), and reflexives (Barlow, Reference Barlow1996). For all these phenomena, researchers have found considerable mismatches between naturally occurring German and English and the type of German or English that is put forward as a model in the examined teaching materials. They have, as a consequence, called for corpus-inspired adjustments in the language-teaching syllabus (particularly as far as selection and progression are concerned) and for revised pedagogical descriptions that present a more adequate picture of the language as it is actually used. A case in point here is the misrepresentation of the functions and contextual patterns of English progressive forms in EFL (English as a foreign language) teaching materials used in German schools. Progressives that refer to repeated actions or events, for example, are considerably more frequent in authentic English than in textbook English, where the common function of repeatedness is neglected and the focus is on single continuous events (cf. Römer, Reference Römer2005, pp. 261–263).

Specialized corpora, for example, those capturing a particular language for special or specific purposes (LSP), can also influence syllabus design for LSP courses. As Gavioli (Reference Gavioli2006) pointed out, in English for specific purposes (ESP), “working out basic items to be dealt with is a key teaching problem” (p. 23). A keyword analysis based on a corpus that contains the specific text or discourse type in question (e.g., English business letters, medical research articles, or newspaper editorials) can help solve this problem and assist teachers in “focus[ing] their efforts in terms of selection of language contents” (Pérez-Paredes, Reference Pérez-Paredes2003, p. 1). An important issue for ESP teachers (who may or, what is more likely, may not be experts in the specific discourse they teach) is that they should give priority to teaching those words and expressions that their learners will need later on to be able to handle texts in their subject area. For instance, having access to a corpus of biology readings and lectures, to give just one example of a science course in English described by Flowerdew (Reference Flowerdew1993), can enable teachers to successfully address this issue and make informed decisions about item and text selection for their course. Another example of how corpora can impact general EAP (English for academic purposes) teaching is Coxhead's (Reference Coxhead2000) Academic Word List (AWL). The AWL, based on a corpus of academic writing, contains the vocabulary items that are most relevant and useful to EAP learners. Simpson-Vlach and Ellis (Reference Simpson-Vlach and Ellis2010) have taken Coxhead's idea from word to phrase level and devised an Academic Formulas List (AFL), which consists of word combinations that occur significantly more often in academic than in nonacademic speech and writing. Both studies (AWL and AFL) take an indirect approach to using specialized corpora in language teaching and contribute to improving the teaching of EAP through informing syllabus design, for example, by helping with the selection of items to be included in the course.

Studies on learner corpora, that is, systematic computerized collections of the language produced by language learners, are also highly relevant for syllabus design (cf. Aston, Reference Aston, Burnard and McEnery2000; Granger, Reference Granger, Granger, Hung and Petch-Tyson2002, Reference Granger and Aijmer2009) since they provide insights on “the needs of specific learner populations” (Meunier, Reference Meunier, Granger, Hung and Petch-Tyson2002, p. 125) and help to test teachers’ intuitions about whether a particular phenomenon is difficult or not (Granger, Reference Granger, Granger, Hung and Petch-Tyson2002, p. 22). Nesselhauf (Reference Nesselhauf2005), for example, identified unidiomatic verb-noun collocations in German learner English (e.g., reach an aim, drive a bike), indicating that combinations such as achieve an aim or ride a bike may deserve particular attention in teaching materials designed for learners with first language (L1) German.

Corpus Research and Designing Teaching Materials

The results of the comparisons between corpora and course books mentioned earlier not only inform the language-teaching curriculum but also help with decisions about the presentation of items and structures in reference works and teaching materials. Corpora research has exerted a huge influence on reference publishing and has led to a new generation of dictionaries and grammar books. Nowadays, “people who have never heard of a corpus are using the products of corpus research” (McEnery, Xiao, & Tono, Reference McEnery, Xiao and Tono2006, p. 97). In the context of ELT, the publications in the Collins COBUILD series constitute a major achievement. Based on real English and compiled with the needs of the language learner in mind, the COBUILD dictionaries, grammars, usage guides, and concordance samplers (see Capel, Reference Capel1993; Carpenter, Reference Carpenter1993; Goodale, Reference Goodale1995; Sinclair, Reference Sinclair1990, Reference Sinclair1992, Reference Sinclair2001) offer teachers and learners more reliable information about the English language than any of the more traditional reference grammars or older non-corpus-based dictionaries. Two major advantages of the COBUILD and other corpus-based reference works for learners (e.g., Biber & Conrad, Reference Biber and Conrad2009; Biber, Leech, & Conrad, Reference Biber, Leech and Conrad2002; Bullon, Reference Bullon2006; Carter & McCarthy, Reference Carter and McCarthy2006; Hornby, Reference Hornby2005; Peters, Reference Peters2004; Rundell, Reference Rundell2007) are that they incorporate corpus-derived findings on frequency distribution and register variation, and that they contain genuine instead of invented examples. Particularly worth mentioning here is the student version of the entirely corpus-based Longman Grammar of Spoken and Written English (Biber et al., Reference Biber, Leech and Conrad2002). The importance of presenting learners with authentic language examples has been stressed in a number of publications (e.g., Fox, Reference Fox and Sinclair1987; Kennedy, Reference Kennedy and Svartvik1992; Römer, Reference Römer, Aston, Bernardini and Stewart2004a, Reference Römer2005; Sinclair, Reference Sinclair1991, Reference Sinclair, Wichmann, Fligelstone, McEnery and Knowles1997). Kennedy (Reference Kennedy and Svartvik1992), for instance, cautioned that “invented examples can present a distorted version of typicality or an over-tidy picture of the system” (p. 366), and Sinclair (Reference Sinclair1991) called it an “absurd notion that invented examples can actually represent the language better than real ones” (p. 5). Thanks to the corpus revolution, language learners can today choose from a range of reference works that are thoroughly corpus-based and that offer improved representations of the language they want to study. While course books and other materials used in the language teaching classroom have long been lagging behind this development and have been rather unaffected by advances in corpus linguistics (at least as far as the English as a foreign language market is concerned), the first attempts are now being made to produce textbooks that draw on corpus research and are fully based on real-life data, that is, on language that has in fact occurred (e.g., Barlow & Burdine, Reference Barlow and Burdine2006; Carter, Hughes, & McCarthy, Reference Carter, Hughes and McCarthy2000; McCarthy, McCarten, & Sandiford, 2005). Some particularly innovative developments are currently taking place in Japan, where Yukio Tono of the Tokyo University of Foreign Studies hosts the world's first corpus-based English conversation program on television (watched by a million Japanese language learners) and publishes a series of accompanying textbooks that have become best sellers in Japan (Tono, Reference Tono, Frankenberg-Garcia, Flowerdew and Aston2011). Each unit of the program focuses on one of 100 keywords (selected from a BNC frequency list) and its most common collocation patterns. Model skits and exercises illustrate the use of these collocations.

Another branch of general corpora research that has exerted some influence on the design of reference works and, to a lesser extent, teaching materials is the area of phraseology and collocation studies. Scholars like Biber, Johansson, Leech, Conrad, and Finegan (Reference Biber, Johansson, Leech, Conrad and Finegan1999); Hunston and Francis (Reference Hunston and Francis2000); Kjellmer (Reference Kjellmer, Aarts and Meijs1984); Lewis (Reference Lewis1993, Reference Lewis1997, Reference Lewis2000); Meunier and Gouverneur (Reference Meunier, Gouverneur, Hidalgo, Quereda and Santana2007); Nattinger (Reference Nattinger1980); Pawley and Syder (Reference Pawley, Syder, Richards and Schmidt1983); and Sinclair and Renouf (Reference Sinclair, Renouf, Carter and McCarthy1988) have emphasized the importance of recurring word combinations and prefabricated strings in a pedagogical context because of their great potential in fostering fluency, accuracy, and idiomaticity. Although corpus-based collocation dictionaries (e.g., Benson, Benson, & Ilson, 2010; Hill & Lewis, Reference Lewis1997; Lea, Reference Lea2002) are available, and although information on phraseology (i.e., about the combinations that individual words favor) is implicitly included in learner dictionaries in the word definitions and the selected corpus examples—and sometimes even explicitly described, for example, in the grammar column in the COBUILD dictionaries and in the COBUILD Grammar Patterns reference books (e.g., the “V over n” pattern for the verb argue; Francis, Hunston, & Manning, 1996, 1998)—such information and exercises on typical collocations are as yet largely missing from language-teaching course books (or they are inadequate; cf. Meunier & Gouverneur, Reference Meunier, Gouverneur, Hidalgo, Quereda and Santana2007). Like Hunston and Francis (Reference Hunston and Francis2000), I see a necessity in and “look forward to [more] information about patterns being incorporated in language teaching materials.” (p. 272)

It has also been demonstrated in recent publications (e.g., Aijmer, Reference Aijmer, Granger, Hung and Petch-Tyson2002; Altenberg & Granger, Reference Altenberg and Granger2001; Herriman & Boström Aronsson, Reference Herriman, Boström Aronsson and Aijmer2009; Leńko-Szymańska, Reference Leńko-Szymańska, Hidalgo, Quereda and Santana2007; Nesselhauf, Reference Nesselhauf2005) how the findings of studies based on learner corpora such as the ICLE can “enrich usage notes” in learners’ dictionaries (Granger, Reference Granger, Granger, Hung and Petch-Tyson2002, p. 24), or how they “can provide useful insights into which collocational, pragmatic or discourse features should be addressed in materials design” (Flowerdew, Reference Flowerdew, Ghadessy, Henry and Roseberry2001, pp. 376–377).

Examples of Direct Pedagogical Corpus Applications

Following Johns's example, a number of researchers have discussed ways in which corpora and concordances can be used by language learners. Bernardini (Reference Bernardini, Kettemann and Marko2002), for example, described the positive effects of what she calls corpus-aided discovery learning with the BNC, and described corpora as “rich sources of autonomous learning activities of a serendipitous kind” (p. 165; see also Bernardini, Reference Bernardini, Burnard and McEnery2000b; 2004a). She envisioned the learner in the role of a “traveller instead of a researcher” (Bernardini, Reference Bernardini, Burnard and McEnery2000a, p. 131; italics in original), and was less “interested in the starting or end point of a learning experience” than in what the learner experiences in between, on her or his journey (Bernardini, Reference Bernardini, Burnard and McEnery2000a, p. 142). Kettemann (Reference Kettemann1995), too, stressed the exploratory aspect of DDL and considered concordancing in the ELT classroom “motivating and highly experiential” for the learner (p. 30).

To give a concrete example of a possible DDL task, learners could be asked to compile concordances of a pair of near-synonyms (such as speak and talk; see concordance samples in Figure 2) and work out the differences in the collocational and phraseological behavior of these words. Further examples of DDL activities are described in Aston (Reference Aston, Wichmann, Fligelstone, McEnery and Knowles1997, Reference Aston, Burnard and McEnery2001, Reference Aston and Lombardo2009); Bennett (Reference Bennett2010); Fligelstone (Reference Fligelstone1993); Gavioli (Reference Gavioli and Aston2001, Reference Gavioli2006); Hadley (Reference Hadley2001); Johns (Reference Johns, Johns and King1991, Reference Johns, Kettemann and Marko2002); Kaszubski (in press); O'Keeffe, McCarthy, and Carter (Reference O'Keeffe, McCarthy and Carter2007); Oksefjell-Ebeling (Reference Oksefjell-Ebeling and Aijmer2009); Reppen (Reference Reppen2010); Sripicharn (Reference Sripicharn, Aston, Bernardini and Stewart2004); and especially in Tribble and Jones (Reference Tribble and Jones1997). Also worth mentioning are some materials and tools for the creation of DDL exercises that are available online, for example, exercises based on MICASE (see http://micase.elicorpora.info/, links to “ESL/EAP Teaching Materials” and “ESL Self-Study Activities”), Tom Cobb's “Compleat Lexical Tutor” (http://www.lextutor.ca/), a tool collection “[f]or data-driven language learning on the web” (Web site blurb), and a module on “Using concordance programs in the modern foreign languages classroom” of the Information and Communications Technology for Language Teachers, http://www.ict4lt.org/en/en_mod2–4.htm.

Fig. 2.

Concordance samples of “speak” and “talk,” based on the spoken part of the BNC.

The DDL method of using learner-centered activities with the teacher as a facilitator of these activities has, up to now, mainly been discussed with reference to ELT and English language corpora. Some of the few studies that deal with other languages include Whistle (Reference Whistle1999) and Chambers (Reference Chambers2005), who report on introducing DDL activities to the teaching of French in order to supplement other CALL (computer-assisted language learning) tasks; Dodd (Reference Dodd, Wichmann, Fligelstone, McEnery and Knowles1997) and Jones (Reference Jones, Wichmann, Fligelstone, McEnery and Knowles1997), who show how corpora of written and spoken German can be exploited “to give students a richer language-learning experience in the foreign language environment” (Dodd, Reference Dodd, Wichmann, Fligelstone, McEnery and Knowles1997, p. 131); and Kennedy and Miceli (Reference Kennedy and Miceli2001, Reference Kennedy, Miceli, Kettemann and Marko2002, Reference Kennedy and Miceli2010) who suggest corpus consultation for learners of Italian.

The Effectiveness of Direct Pedagogical Corpus Applications

Advantages of corpora work with learners have been suggested by scholars like Sinclair (Reference Sinclair, Wichmann, Fligelstone, McEnery and Knowles1997), who noted that, for the learner, “[c]orpora will clarify, give priorities, reduce exceptions and liberate the creative spirit” (p. 38). Likewise, many researchers and teachers in the TaLC tradition are convinced that DDL can empower learners to find out things for themselves, that a data-driven approach raises the learners’ language awareness, and that corpora have great pedagogic potential. The effectiveness of DDL has actually been indicated in studies on the teaching and learning of grammar and vocabulary by Boulton (Reference Boulton2009); Cobb (Reference Cobb1997); Cresswell (Reference Cresswell, Hidalgo, Quereda and Santana2007); Granath (Reference Granath and Aijmer2009); Johns, Hsingchin, and Lixun (Reference Johns, Hsingchin and Lixun2008); and Yoon (Reference Yoon2008). As Yoon (Reference Yoon2008) observed in a study on the influence of corpora use on L2 academic writing, “students assumed more responsibility for their writing and became more independent writers” (p. 31). On a similar note, Johns et al. (Reference Johns, Hsingchin and Lixun2008) found that “CALL activities [including DDL] did have a positive impact on students' English learning” (p. 503), and Boulton (Reference Boulton2009) noted that in his empirical study “corpus samples led to more successful results than traditional pedagogical resources” such as bilingual dictionaries or usage manuals (p. 50). Concordancing has not only been shown to be a useful way “to mimic the effects of natural contextual learning” (Cobb, Reference Cobb1997, p. 314), but researchers have recently also highlighted its use and usefulness in translation teaching (Bernardini, Reference Bernardini and Malmkjaer2004b; Kübler, Reference Kübler, Frankenberg-Garcia, Flowerdew and Aston2011) and for error correction in foreign or second language writing (cf. Bernardini, Reference Bernardini and Malmkjaer2004a; Chambers, Reference Chambers2005, Reference Chambers, Hidalgo, Quereda and Santana2007; Gaskell & Cobb, Reference Gaskell and Cobb2004). These descriptive empirical studies demonstrate that corpora nicely complement existing reference works, and that they may provide information that a dictionary or grammar book may not provide. They are based on experiments in which groups of students use corpora, corpus-derived exercises, and/or concordances in the classroom, while control groups of students use traditional teaching materials.

Although according to Granath (Reference Granath and Aijmer2009), “advanced students definitely benefit from working with corpora” (p. 59), it is not clear whether this is also the case for beginning learners of a foreign language. Dealing with concordances based on large collections of native-speaker conversations, newspaper texts, research articles, or novels that contain many unknown words and complex sentence constructions may be too challenging a task for beginners, and not all learners may feel comfortable working with computers in language learning. A further limitation may be that it initially takes time for teachers to familiarize themselves and/or their students with online corpora or offline corpus tools. However, once they have been given basic training in corpus use, teachers should be able to create simple concordance exercises tailored to their learners’ needs and access information about collocation and language patterning. Teachers could, for example, access the COCA interface (http://www.americancorpus.org/) to search for a word or phrase and retrieve concordance samples and lists of collocations for this word or phrase. As a recent survey on teachers’ needs has shown (see Römer, Reference Römer2009a), teachers of a language that is not their L1 often require native-speaker advice on language points. Computer corpora that have been described as “tireless native-speaker informant[s], with rather greater potential knowledge of the language than the average native speaker” (Barnbrook, Reference Barnbrook1996, p. 140), can offer help in such situations.