Presenters: Kristin Terrill and Dr. Elena Cotos

Abstract: Learning technologies have been implemented in higher education worldwide. Technology-based writing tools provide students automated feedback, mostly on grammar, style and mechanics. Few of those tools, however, were designed for graduate students producing advanced academic genres (e.g., AcaWriter (Shibani, Knight, & Buckingham Shum, 2019) and Thesis Writer (Kruse & Rapp, 2019)). Even fewer were grounded in a needs analysis, which is considered mandatory in English for Academic Purposes (Brown, 2016). 

This paper reports a needs analysis motivated by concerns that published theses and dissertations contain issues that detract from quality and impact. Previous research shows that novice scholarly writers struggle with both local and global aspects of academic writing. Local problems include inaccurate grammatical forms, inappropriate vocabulary choices, and mechanics issues (Casanave & Hubbard, 1992; Cooley & Lewkowicz, 1995, 1997; Surratt, 2006). Global issues include weak arguments, uncritical evaluation, unsupported claims, lack of structure, inappropriate emphasis, and organization problems (Alter & Adkins, 2006; Dong, 1998; Lim, 2012; Flowerdew, 1999; Thompson, 1999; San Miguel & Nelson, 2007). The present study aimed to identify and describe such problems in theses and dissertations to inform a writing analytics tool for pre-screening publication submissions. 

Corpus-based research has demonstrated potential for digital writing technologies (e.g., Research Writing Tutor (Cotos, 2016)). Therefore, our study involved analyzing a corpus of 126 randomly selected theses and dissertations published in ProQuest (dissertations repository for the U.S. Library of Congress) between 2015-2017. All texts were written by graduates of one large U.S. university; they were representative in terms of discipline (58 disciplines), student status (49 masters; 77 doctoral), and IMRD/C (introduction-methods-results-discussion/conclusion) structure. Four coders analyzed the corpus using an inductively developed framework. The coded data enabled quantification and comparison across student status, discipline, and text section. The results indicate that the prevalence and severity of local and global issues depend on these variables. These findings point to target needs, which supports our goal of designing a writing analytics tool for students writing their thesis or dissertation. We anticipate that this application could be integrated into ProQuest more broadly. The presentation will conclude with implications for the tool’s design and implementation.