Authorship attribution is one of the largest areas of Forensic Linguistics. It entails determining who wrote, or more commonly, ruling out who wrote a text when authorship is unclear. Linguists assume that each particular choice the writer makes as a whole will enable identification as authors are consistent in their choices. Linguists have three main problematic scenarios when attempting authorship attribution; there is no candidate set and a profile is required, there are many candidates for a limited sample and verification where one determines whether the suspect is the author or not. In regards to whether a communication was written by the suspect or deceased, we would need to focus upon the third scenario. Documents that would entail authorship dispute regarding a deceased individual would include wills, last testaments and suicide notes.
In the case of a suicide note being left behind the linguist has several issues to consider. Most importantly, why the deceased may have committed suicide. Normally it has little relevance to the linguist why a person commits a crime or does something but knowing the reasons why will help distinguish genuine from simulated and it is almost always the first question relatives ask. In order to discover why, it is useful to know what friends and family feel after initial grief. Common views are that the person was crazy, a coward or a loser. This popular view of suicide helps the linguist to an extent as a fake suicide note is more likely to reflect popular attitudes instead of how the person actually feels.
One of the main approaches to the methodology of authorship attribution is the Unitary Invariant Approach which was pioneered by Mendenhall (1887) who sought to distinguish the literary style of those such as Bacon and Shakespeare. This attempt at a scientific approach was based on the assumption that a unique curve would appear which showed the relationship between word length and relative frequency and could be used to identify an author where none was known. By the twentieth century those such as Zipf (1932) and Yule (1944) helped to update this method through the suggestion that a feature or marker such as sentence length could help identify a possible author. However, they admit themselves that this method using statistics is unreliable (Sichel 1986; Burrows 1992; Grieve 2007) especially for the much shorter texts that many Forensic Linguists have to work with.
Another key moment that revolutionised authorship attribution was Mosteller and Wallace’s (1964) work on the Federalist papers. They argued that stylometrics (which combined information from textual clues) was more successful in authorship attribution. By analysing the frequencies of functioning grammar words it yielded a reliable method which opened up the field to new textual features and modelling techniques. Principal components analysis (PCA) furthered this idea and found that “if good separation is seen between...