A corpus-based study of academic-collocation use and patterns in postgraduate Computer Science students’ writing

Collocation has been considered a problematic area for L2 learners. Various studies have been conducted to investigate native speakers’ (NS) and non-native speakers’ (NNS) use of different types of collocations (e.g., Durrant and Schmitt, 2009; Laufer and Waldman, 2011).These studies have indicated that, unlike NS, NNS rely on a limited set of collocations and tend to overuse them. This raises the question: if NNS tend to overuse a limited set of collocations in their academic writing, would their use of academic collocations in a specific discipline (Computer Science in this study) vary from that of NS and expert writers? This study has three main aims. First, it investigates the use of lexical academic collocations in NNS and NS Computer Science students’ MSc dissertations and compares their uses with those by expert writers in their writing of published research articles. Second, it explores the factors behind the over/underuse of the 24shared lexical collocations among corpora. Third, it develops awareness-raising activities that could be used to help non-expert NNS students with collocation over/underuse problems. For this purpose, a corpus of 600,000 words was compiled from 55 dissertations (26 written by NS and 29 by NNS). For comparison purposes, a reference corpus of 600,269 words was compiled from 63 research articles from prestigious high impact factor Computer Science academic journals. The Academic Word List (AWL) (Coxhead, 2000) was used to develop lists of the most frequent academic words in the student corpora, whose collocations were examined. Quantitative analysis was then carried out by comparing the 100 most frequent noun and verb collocations from each of the student corpora with the reference corpus. The results reveal that both NNS (52%) and NS (78%) students overuse noun collocations compared to the expert writers in the reference corpus. They underuse only a small number of noun collocations (8%). Surprisingly, neither NNS nor NS students significantly over/underused verb collocations compared to the reference corpus. In order to achieve the second aim, mixed methods approach was adopted. First, the variant patterns of the 24 shared noun collocations between NNS and NS corpora were identified to determine whether over/underuse of these collocations could be explained by their differences in the number of patterns used. Approximately half of the 24 collocations identified for their patterns were using more patterns including (Noun + preposition +Noun and Noun + adjective +Noun) that were rarely located in the writing of experts. Second, a categorisation judgement task and semi-structured interviews were carried out with three Computer Scientists to elicit their views on the various factors likely influencing noun collocation choices by the writers across the corpora. Results demonstrate that three main factors could explain the variation: sub-discipline, topic, and genre. To achieve the third pedagogical aim, a sample of awareness-raising activities was designed for the problematic over/underuse of some noun collocations. Using the corpus-based Data Driven Learning (DDL)approach (Johns,1991), three types of awareness-raising activities were developed: noticing collocation, noticing and identifying different patterns of the same collocation, and comparing and contrasting patterns between NNS students’ corpora and the reference corpus. Results of this study suggest that academic collocation use in an ESP context (Computer Science) is related to other factors than students’ lack of knowledge of collocations. Expertness, genre variation, topic and discipline-specific collocations are proved important factors to be considered in ESP. Thus, ESP teachers have to alert their students to the effect of these factors in academic collocation use in subject specific disciplines. This has tangible implications for Applied Linguistics and for teaching practices.