In this paper a novel approach to authorship attribution is suggested, specifically relating to conceptual metaphor as exemplified in Lakoff (1994).
In the history of modern authorship attribution there has probably been no more widely analysed and measured set of texts, with the probable exception of Shakespeare, than the Federalist Papers, a set of 85 documents prepared for the citizens of New York in the late 1780’s to persuade them to adopt a federal system of government rather than allowing the then 13 states to fracture into confederacies or separate political entities. Writing these texts was the brainchild of Alexander Hamilton, a native of the West Indies who had gone to America as a young student of King’s College (now Columbia University). It was to be a mammoth undertaking and initially three men were to be involved in their creation: Alexander Hamilton himself, James Madison (later the 4th President of the United States) and John Jay, a prominent lawyer. All of them were central figures in the political and legal establishment of their day. Traditionally, Hamilton is seen as the author of 51 of the papers, Jay contributed only 5 of them (due to ill health) and Madison 14. Three of the papers are jointly attributed to Hamilton and Madison, and twelve are of disputed authorship — the work of either Hamilton or Madison. Most of the authorship attribution tests which have been carried out on these twelve papers have attributed them to James Madison. The most famous analysis of their authorship was conducted by Mosteller and Wallace (1964).
My objection to many of the authorship tests carried out on the Federalist Papers, including that of Mosteller and Wallace, is threefold: firstly, the means of analysis is not linguistic in most cases, but statistical. Doubtless, Mosteller and Wallace’s Bayesian analysis is brilliantly simple, but it cannot be described as having any foundation in linguistic theory. It merely computes frequencies of a set of function words from each of the candidate authors without saying why those particular function words should have any linguistic bearing on the delicate quality of authorship.
My second objection to Mosteller and Wallace (hereafter MW) is the source of their texts. There are several editions of the Federalist Papers and, as MW themselves admitted, they made many editorial corrections to the papers before testing them. Rudman notes that the ‘little book of corrections’ MW refer to in their paper, lists over a thousand such corrections but, sadly, this ‘book of corrections’ has never been made public and so we do not know what those alterations consisted of.
My third objection to MW is that, like many computational analyses, little or no attention is paid to the actual language of the texts. For all I know, tests like MW’s could be applied to a description of surgical procedures or an opera programme with equal results: I am cautious of rendering decisions on authorship by quantification alone, unless the method of quantification has some foundation in linguistic theory. However, in the case of a large corpus of long texts I am prepared to be persuaded that some statistical tests may be effective, providing that proven statistical methods are used. I say this because the fact that many other studies have, apparently independently, confirmed Madison as the author of the disputed papers, must engender some respect for the methods used, even if, as linguists, we may feel that the case has not been made according to linguistic principles. This is not to suggest that we should rely absolutely on such methods, but that it might be possible to include them in our authorship attribution arsenal, with the proviso that we should not neglect the development of linguistic methods in the task of authorship attribution.
In this latter regard, I would like to propose a broadly cognitive linguistic approach, based largely on Lakoff’s notion of conceptual metaphor. In order to set forth how I have arrived at this position, I will first describe some aspects of intertextuality and how these may impinge on the question of authorship.