Dimensions of variation across Internet registers

Abstract

Abstract This paper presents a study that sought to identify the dimensions of variation underlying a corpus of Internet texts, using Biber’s (1988) multi-dimensional (MD) analysis framework. The corpus was compiled following the method proposed by Biber (1993), according to which the size of each register subcorpus should be determined based on the linguistic variation across the texts. The corpus was tagged using the Biber Tagger and the features were counted and submitted to a factor analysis, which suggested three factors. The factors were interpreted as three dimensions of variation: involved, interactive discourse versus informational focus; expression of stance: interactional evidentiality; and expression of stance: interactional affect. The amount of register variation captured by the register distinctions on the dimensions ranged from 8.7% to 57.1%. Dimension 1 corroborate the oral/involved versus literate/informational distinction defined in previous MD studies of non-Internet registers, whereas Dimensions 2 and 3 highlight the important role played by stance in social media.