Corpus-aided language learning

A ‘corpus’ is a large collection or database of machine-readable texts involving natural discourse in diverse contexts (Bernardini 2000). Such discourses can be spoken, written, computer-mediated, spontaneous, or scripted and may represent a variety of genres (for example everyday conversations, lectures, seminars, meetings, radio and television programmes, and essays). Some readily available corpora include the British National Corpus (BNC, http://www.natcorp.ox.ac.uk), which contains 100 million words from written and spoken language in a variety of contexts, the Michigan Corpus of Academic Spoken English (MICASE, http://micase.elicorpora.info), which features 1.8 million words of speech in various academic contexts, and the Corpus of Contemporary American English (COCA), with 410 million words (http://www.americancorpus.org).1 Although corpus linguistics (i.e. computer-assisted analysis techniques for studying texts) is a young specialization, its usefulness in teaching and learning has received growing attention and recognition (for example Hunston 2002; Sinclair 2004; Conrad 2005; O’Keeffe, McCarthy, and Carter 2007; Bennett 2010; Reppen 2010). In particular, researchers have identified corpus data as resources that provide descriptive insights relevant to how people use language and as tools that enable students and instructors to analyse both how people use different language forms at various levels of formality and how language fulfills multiple speech functions across contexts. Corpus data suggest that individuals often do not use language as specified in grammar books and that word meanings vary across contexts and users (Biber and Reppen 2002). Over the past ten years, a growing number of studies have shown how learners can use corpus data to further their language learning (see Hunston op.cit.; Boulton 2010). Numerous corpus linguists (for example Gavioli and Aston 2001) have pointed out that learning activities centred on analysing corpus data are consistent with current principles of language-learning theory, that is students develop more autonomy when they receive guidance about how to observe language and make generalizations. Such activities promote noticing and grammatical consciousness raising (Schmidt 1990), which can enhance second language learning and development. Despite the growing interest in corpora and corpus-aided learning, however, many teachers believe that incorporating corpora into their teaching would be too technically challenging or time consuming (Boulton 2010). Yet, while some researchers have suggested substantial training is necessary (for example Estling Vannestål and Lindquist 2007), others have provided evidence that only a minimal amount of training is needed (for example Boulton 2008). Some have also recommended using paper-based materials generated from corpora as a viable alternative to accessing corpora via computers (Boulton 2010). A key pedagogical approach for using corpora in language teaching and learning is ‘data-driven learning’ (DDL), which emerged in the mid-1980s. DDL was defined as ‘the use in the classroom of computer generated concordances to get students to explore the regularities of patterning in the target language, and the development of activities and exercises based on concordance output’ (Johns and King 1991: iii). As Johns (1994: 297) stated, ‘what distinguishes the DDL approach is the attempt to cut out the middleman as far as possible and to give direct access to the data so that the learner can take part in building up his or her own profiles of meaning and uses’. Furthermore, corpus data ‘[offer] a unique resource for the stimulation of inductive learning strategies—in particular the strategies of perceiving similarities and differences and of hypothesis formation and testing’ (ibid.). By extension, the corpus-aided discovery learning (CADL) approach entails encouraging learners to take the role of language researchers by systematically engaging in discovery learning (Gavioli 2000) and in learning how to learn through observations, analyses, interpretations, and presentations of language-use patterns in corpus data. In the CADL approach, learning about language use is driven by a process of enquiry that works toward understanding or problem solving, and corpora are used as mediational tools (Vygotsky 1978) rather than as the basis for language teaching and learning. Furthermore, instructors adhering to the CADL approach play a critical role in facilitating or guiding the process of discovery, which depends on the learners’ needs, stages of learning, and levels of proficiency. Researchers have generally agreed that corpus data enrich our understanding of language use and are an important resource for language teaching and learning. The use of corpora in language teaching is not without controversies, however. Among the debates featured in Seidlhofer (2003), for example, some scholars have advocated using ‘real examples only’ in the classroom (for example Sinclair 1997), while others, in contrast, wonder whether the discourse in corpora, taken out of its original context, can still be considered ‘authentic’, ‘real’, or ‘natural’, thereby questioning the efficacy of analysing displaced language that may not be relevant to learners’ linguistic and sociocultural contexts. In response to Widdowson’s (1998) remark that corpora may provide samples of genuine language produced by language users with real communication goals but do not necessarily guarantee that learners can participate in discourse in ways that lead to learning, researchers such as Gavioli and Aston (op.cit.) note that learners can still ‘authenticate’ language samples by adopting an observer’s role to critically analyse the data, which will raise their awareness of lexical, grammatical, and textual issues as they restructure their views about language use in real situations. Similarly, Carter (1998: 50–1) argues that while ‘real’ English from corpora can be ‘unrealistic’ for classroom instruction and thus modified language used in the classroom that is based on learners’ needs and levels might be more ‘pedagogically viable and realistic’, learners should be provided with opportunities to develop a ‘feel’ for the language through corpus data. The validity of analysing corpora to capture language use across seemingly limitless contexts or to describe the workings of ‘real English’ around the world has also been questioned. Some scholars point out that communicative contexts are not restricted to native speaker discourse, and, as such, language teaching should not be based simply on descriptive facts generated from largely native speaker-oriented corpora (Prodromou 1996).2 Despite these debates, technological advancements have undoubtedly enhanced language learners’ and instructors’ access to corpora, and the plethora of articles and books written for language-teaching researchers and practitioners published during the past five years suggest that attention to and interest in using corpora for teaching and learning purposes will continue for the foreseeable future.