Conversation text types: A multi-dimensional analysis

Multi-dimensional (MD) analysis is a methodological approach that applies multivariate statistical techniques (especially factor analysis and cluster analysis) to the investigation of register variation in a language. The approach was originally developed to analyze the full range of spoken and written registers in a language. Early studies focused on English register variation (Biber 1985, 1986 and 1988), while later studies have applied the same approach to Somali, Korean, Tuvaluan, Taiwanese, and Spanish. Surprisingly, these studies have found some striking similarities in the underlying ‘dimensions’ that distinguish among spoken and written registers in these diverse languages. It is even more surprising that MD studies of restricted discourse domains have also uncovered dimensions that are similar in linguistic form and function to the more general studies of register variation. The present study presents an MD analysis of a single register: conversation. Three primary dimensions of variation are identified, and then cluster analysis is used to distinguish among six conversation text types. The dimensions and text types are interpreted in linguistic and functional terms. The author’s expectations were that a unique set of dimensions would emerge to characterize the variation among conversational texts. Instead, the three dimensions identified here turn out to be closely related to dimensions identified in previous analyses of general register variation. Taken together with previous studies, the present study of conversation raises the possibility of universal dimensions of variation.