Capturing Herder: a three-step approach to the identification of language ideologies using corpus linguistics and critical discourse analysis

Abstract

Recent lexical approaches to the identification of language ideologies focus on the application of quantitative corpus-linguistic techniques to large data sets as a way to minimise researcher inference and ensure more objective sampling methods, replicability of analytical procedures, and a higher degree of generalisability ( Fitzsimmons-Doolan, 2014 ; Subtirelu, 2015 ; Vessey, 2017 ; Wright and Brooks, 2019 ; and McEntee-Atalianis and Vessey, 2020 ). Based on two comprehensive, specialised research (11.6 million words) and comparator (22.4 million words) newspaper corpora, this study offers an examination of the effectiveness of the multivariate and univariate statistical techniques, and proposes a three-step approach whereby corpus linguistics and critical discourse analysis are combined to identify ( 1) thematic and ( 2) ideological discourses (cf. ‘d’/’D’ discourses; Gee, 2010 ), and ( 3) language ideologies. In contrast to recent contributions, it is argued that item frequency is not necessarily a reliable or effective indicator of language ideologies but, rather, of language-related discourses which can be examined for implicit and explicit language-ideological content. A combination of multivariate and univariate statistical techniques, and the three-step approach are shown to be a highly effective methodological solution for synchronic and diachronic language ideology and discourse research based on topically/discursively heterogeneous corpora.