The microanalysis of style variation

Abstract

Computational stylistics has often analyzed variations in the style of a single author or text, including chronological change, the dialogue or narration of multiple characters or narrators in a single novel, and other perceived shifts in style. Here, I examine intra-textual style variation from a different perspective, and suggest circumstances under which it is appropriate and beneficial to omit some parts of texts from an analysis to eliminate sources of predictable and intrusive variation that would render other kinds of variation or consistency of style invisible. In other cases it may be appropriate to randomize a text to mask a kind of variation that would otherwise disrupt an analysis. One way to do this is by sorting the lines of the text in random order before analyzing it in sections. In analyses in which texts of very different lengths must be compared, it is also beneficial to randomize the longer parts, cut them all to approximately the length of the shortest one, and use the equalized amounts of text to create a word frequency list that is not biased in favor of the longer parts. These techniques can lead to a deeper understanding of style variation.