Attributing Documents to Candidate Authors Using Idiolectal Co-selection and the Web

Abstract

In this chapter, empirical research is described where a Questioned Document is compared to the known writings of three candidate authors, with the aim of establishing whether idiolectal co-selections are a useful marker of authorship and whether the web can be used to generate reliable evidence. Lexical items are identified which are not core words, derivatives of core words or subject-specific words, and which are argued to constitute a set of idiolectal co-selections for each author. Results show that idiolectal co-selections hold some potential as a marker of authorship for prolific authors on the web, but reliability between search engines is low, as is reliability over a period of one week. Attempting to repeat the results after a period of ten years revealed that web searches cannot be repeated exactly, which raises issues for the forensic context.