ABSTRACT

This chapter describes the theoretical framework used by the present work – the multi-feature, multidimensional (MM) framework for studying register variation developed by Biber; it then introduces the statistical method of Correspondence Analysis, which is different from the Factor Analysis used by most MM-style work. Previous research has mostly focused on isolated features, such as word length, lexical and syntactic choices, and, especially, 'colloquial versus literary' doublets. Correspondence Analysis offers the advantages over Factor Analysis. Its greatest appeal lies in its use of the intuitive bi-plot visualization of dimensions, which can help in detecting structural relationships among the variables and also aides in interpretation of the dimensions. For researchers familiar with Factor Analysis, a notable difference exists in how data are coded with Correspondence Analysis, at least in its scheme in Correspondence Analysis (SPSS) implementation. The current version of SPSS (v23) supports both English and Chinese for feature/register labels.