ABSTRACT

This chapter discusses methodological issues relating to keyness analysis, and addresses a number of interconnected themes. It offers awareness of relevant methodological choices and their implications, and also addresses related misconceptions and resulting practices, particularly regarding the selection of linguistic units, appropriate metrics, and thresholds of frequency, effect-size and statistical significance. The chapter also discusses the pervasive partiality in keyness analysis, as the vast majority of keyness studies focus on difference, at the expense of similarity. It describes the tension between objectivity and subjectivity in relation to methodological choices, and problematises the frequent conflation of quantitative analysis and objectivity. In adopting a statistical significance score as the indication of keyness, WordSmith Tools conformed to contemporary widespread practice in disciplines employing quantitative analyses. In fact, it is not unlikely that the wording of the definition of keywords was influenced by the choice of the particular statistical significance metric in Wordsmith Tools, log likelihood.