CONCORDANCE-BASED FEEDBACK FOR L2 WRITING IN AN ONLINE ENVIRONMENT

Abstract

Data-driven learning is a sub-discipline of corpus linguistics that makes use of the analyses and tools of corpus linguistics in foreign and second language classroom (Johns, 1991; Johns & King, 1991). With this approach, learners become researchers rather than passive recipients of language rules (Johns, 1991). This study was an investigation of the impact of this approach as a form of written corrective feedback for in-service teachers of English participating in an online writing course at a teacher training institute in Japan. Data-driven learning is commonly utilized in conventional, face-to-face classrooms, or computer lab settings in which there is close direction from the instructor on how to interpret the output of a corpus query. The purpose of this study was to investigate how data-driven learning can be implemented in a blended online environment by providing training to develop the participants’ corpus competence (Charles, 2011; Flowerdew, 2010), which is defined as the ability to interpret data obtained from querying a corpus. This competence has been associated with becoming familiar with corpus methods, which include interpreting concordances, and in turn can aid in accurately repairing writing errors. This training, while initially presented in a face-to-face session at the beginning of the course, was sustained with support from resources on the course’s Moodle website and my comments in Microsoft Word documents. In addition, I applied a fine-grained approach to the analysis of the to examine the quality of participants’ interpretation of concordances. The mixed method triangulation convergence design (Creswell & Plano Clark, 2007, 2011) used in this study was based on data from four sources to examine the effectiveness of data-driven learning in an online environment as well as to observe how the participants interpreted concordances. One data set involved an analysis of the participants’ responses in drafts of their own writing to concordance-based feedback. The participants were given a prefabricated concordance, which was a concordance I generated. That concordance was attached to an error in the participants’ document and the participants used the information provided by the concordance to repair their writing error. The resulting data set, which contains the concordance, along with before and after comparisons of the writers’ repairs, shows how the participants’ interpretations of concordances aided the repairs. With the evidence of several trials over the course of four writing assignments, it was possible to see how the participants used the supplied concordance to repair their writing errors and in turn revealed their degree of corpus competence. A second data set obtained from think-aloud protocols from select participants was utilized to reveal how they interpreted the concordance during an error-repair task. This data revealed what kind of thought processes or noticing that occurred in this task. A third piece of evidence was derived from data obtained from the Moodle website via log files and other resources such as online documents and training quizzes. The purpose was to document which resources the participants accessed relating to data-driven learning training to investigate if those resources aided in their development of corpus competence. The fourth piece of evidence was a quiz developed online to compare the participants with a standard set of items. The quiz was used to investigate which participants successfully or unsuccessfully interpreted the concordances. This instrument, which was analyzed with the Rasch model, allowed for further comparison between the participants’ skill of interpreting concordances. These four data sources were triangulated and in the final analysis cross-referenced to examine how data-driven learning can be successfully applied in a blended online learning environment and how the training of corpus competence aided the learners in interpreting the concordances.