From ‘our methods apply equally well’ to ‘the model does a very poor job’: a corpus-based study of academic value-marking

Disciplinary discourse reflects and reinforces values whose investigation can yield textual evidence of the qualities or aspects of reality regarded as desirable or undesirable by its members. Scholars challenge and (re)negotiate knowledge claims through texts that signal how “the values of academic communities are articulated in discourse meanings” (Hyland 1997: 20). Successful communication depends on the participants’ ability to establish a common ground based on a shared of what is good or bad, suitable or unsuitable, affective or ineffective, etc. But while the speech acts realising evaluation have been extensively researched, little is known of the underlying values they encode. This paper illustrates the methods employed in a recent study of academic writing based on a 1 million-word corpus of English research articles from ten different disciplinary areas (anthropology, biology, computer science, economics, engineering, history, mathematics, medicine, physics and sociology). Starting from candidate items with 100+ wordlist occurrences, the most frequent value-marking categories were investigated using a combination of automated and manual techniques; four of these categories (relevance, size, novelty and goodness) were then singled out for systematic investigation, based on lexical sets and word groups including synonyms and antonyms from different parts of speech. The results show how scholars from different disciplines draw on a repertoire of conventional, largely unqualified, axiological meanings instrumental to the production of knowledge claims in their field. This study specifically targets the axiological variables embdeed in explicitly evaluative speech acts. It employs tools consistent with earlier efforts (e.g. Biber et al. 2004) combining corpus-linguistic and discourse-analytic metho¬d¬o¬lo¬gies for the analysis of lexical variation in academic genres. While the detailed findings are described in Giannoni (2010), the focus here is on the challenges inherent in such an approach: the conventional criteria used for grouping and selecting disciplinary fields; the difficult balance between automated text processing and manual investigation; the difference between evaluative speech acts and parameters of value; insights and directions for further corpus-based work. Despite its mundane appearance, the significance of such data should not be understimated as “every act of evaluation expresses a communal value-system, and [...] this value-system in turn is a component of the ideology which lies behind every text” (Thompson & Hunston 2000: 6)