Anonymity and Imitation in Linguistic Identity Disguise

Abstract

Authorship attribution can be highly accurate, but most techniques are based on the assumption that authors have not attempted to disguise their writing style. Research has found that when writers had deliberately altered their style, commonly used authorship analysis techniques only performed at the level of random chance. This is problematic because many forensic authorship cases investigate documents where it is believed that an author has tried to impersonate somebody else for criminal purposes, and has attempted to adapt their writing style to do so. This study uses a corpus of scripts from the BBC drama, The Archers, to explore how authors write different characters’ voices. Scriptwriters need to adapt their writing style to create the different characters’ dialogues, and this fictional identity disguise is used as a proxy to examine authorship analysis techniques in forensic linguistics. The thesis begins with a literature review exploring the nature of linguistic identity and literary characterisation. It considers the advantages and disadvantages of using fictional data to address forensic problems. There are three main studies: firstly, a quantitative analysis comparing inter-author consistency and variation of authorship analysis features; the second study is a qualitative, stylistic analysis of characterisation, exploring lexical choice, use of dialect, and (im)politeness strategies. The third study is a corpus analysis of the different pragmatic functions of shared lexical tokens. The studies showed that as writers adapted their linguistic style to create different characters, results for commonly-used attribution techniques were observably affected. Some linguistic identities were more distinctive than others, and some authors were more clearly identifiable than others. At a pragmatic level, authors showed more inter-character consistency, and a reduced ability to anonymise their own linguistic traits. This reinforces the importance of investigating linguistic identity disguise at higher levels of language analysis, in addition to lower-level, structural features.