Authorship Attribution of Canonical Texts

Thirteen New Testament letters bear Paul's name, but scholars have long distinguished a core of undisputed letters from several whose authorship is contested on historical, theological, and stylistic grounds. This gives stylometry an unusually clean problem: a substantial body of securely attributed text to define Paul's style and a set of questioned letters to test against it. Savoy's 2019 study exploited exactly this structure, finding that the corpus clusters into groups in which the contested letters separate from the undisputed ones, quantitative support for conclusions that historical critics had reached by other means.

The impostors method addresses open-set verification, where the true author of a disputed text might not be among the candidates. Instead of simply picking the nearest candidate, it compares the disputed text against the claimed author and against a large set of unrelated impostor authors, repeatedly and over randomized subsets of features. The claim is supported only if the text is consistently closer to the candidate than to the impostors. This prevents the common error of forcing every document onto whichever candidate happens to be nearest, and it yields a calibrated confidence score rather than an unconditional verdict.

Yes, and this is the chief caution in scriptural attribution. A single author writes differently across genres, occasions, and stages of life, and ancient letters were often dictated to a secretary or shaped by a co-sender, any of which can move the stylistic profile. Heavy later editing and, above all, working from translation rather than the original language can also distort the signal. For these reasons careful studies validate on known-author texts, use verification rather than mere nearest-match, and present stylometric findings as one strand of evidence to be weighed with historical and theological analysis, not as a standalone proof of forgery.