Investigating macroscopic textual variation through multifeature/multidimensional analyses

This paper distinguishes between two approaches to the analysis of linguistic variation among texts: microscopic and macroscopic. Most previous text analyses have used a microscopic approach, considering in detail the communicative functions of individual linguistic features in particular texts. Macroscopic analysis, on the other hand, sets out to define the underlying parameters of textual variation, or the textual ‘dimensions’, within a set of text types. In the present paper, the different goals and strengths of these two approaches are discussed, followed by a detailed presentation of the macroscopic approach to the analysis of textual variation. The paper argues that only a multifeature/multidimensional approach is adequate for macroscopic analyses. The primary analytical tool in this approach is factor analysis, with explicit emphasis placed on the selection of a broad range of linguistic features and texts for the analysis. Factor analysis statistically groups together linguistic features that cooccur with a high frequency in texts, and these groups of cooccurring features can be interpreted as textual dimensions. The paper describes in detail the use and interpretation of factor analysis in macroscopic text analyses and illustrates this technique through previous research on the linguistic relations among spoken and written text types.