Register variation in malicious forensic texts

Abstract

The study reported here examines a corpus of 104 authentic malicious forensic texts for register variation. A malicious forensic text is defined in this paper as a text that is threatening, abusive or defaming and that constitutes evidence for a forensic case. This corpus was firstly tagged with a set of situational parameters and then analysed using the same multidimensional model introduced in Biber (1988; 1989). The results of the study indicate that malicious forensic texts, similarly to non-malicious professional letters, are on average instances of the Involved Persuasion text type, which is characterised by linguistic features overtly expressing modality. The results also confirm that threatening texts tend to use more modal verbs than non-threatening texts. Furthermore, the personal knowledge between interactants was found to highly influence the level of information density of the texts, while the narrativity level of malicious texts was found to be affected by whether the text contains harmful content directed to the addressee or to a third party. These findings can inform and improve the authorship analysis of malicious texts and increase our understanding of the creation of language crimes.