The Unabomber Case: Stylometric Attribution at Scale

The FBI's Linguistic Analysis Unit began building a profile of the Unabomber from his first communications in 1978 and refined it over seventeen years. The Manifesto, when it arrived in 1995, gave analysts vastly more material than all previous letters combined. The LAU profile noted that the author used formal, educated prose with precise vocabulary, referenced academic disciplines accurately, showed a strong preference for specific rhetorical constructions, and used a number of idiosyncratic phrasing choices that recurred consistently across the text.

One of the most cited markers was the phrase 'feelings of inferiority', which the author deployed repeatedly in a sociological argument about the psychological damage caused by industrialisation. The phrase is not especially rare, but its frequency and specific usage in the Manifesto matched closely with Kaczynski's private journals recovered after the arrest. Other recurring features included 'cool-headed logician', 'clearly' used as an emphatic discourse marker in specific syntactic positions, and the British spelling 'analyse' in a text that was otherwise written in American English.

Post-arrest analysis by both FBI linguists and independent academics compared the Manifesto systematically with two sets of Kaczynski's known writings: his private journals (seized at the cabin and covering decades of his adult life) and his published academic papers from his time as an assistant professor of mathematics at UC Berkeley in the late 1960s.

• Phrase-level matches: specific multi-word expressions that appeared in both the Manifesto and the journals at rates well above what chance occurrence would predict.
• Syntactic patterns: a strong preference for certain subordinate clause structures and a habitual placement of adverbial modifiers that appeared consistently across all Kaczynski-authored text types.
• Lexical idiosyncrasies: unusual word choices ('chipping away at', 'cool-headed', 'attain') used in the same metaphorical contexts across documents separated by years.
• Spelling and punctuation habits: 'analyse', consistent comma splice in specific constructions, and a preference for hyphenated compound modifiers in positions where most American writers do not hyphenate.

Don Foster was an English professor at Vassar College who had built a reputation for attribution work on literary texts, including the disputed poem 'A Funerall Elegie' and, notably, the early identification of Joe Klein as the anonymous author of Primary Colors. After the Manifesto was published in 1995, Foster analysed it and submitted his findings to the FBI, noting a profile consistent with an academic in the physical sciences or mathematics, likely educated in the 1950s or 1960s at a research university, with specific stylistic characteristics he documented.

Foster's contribution to UNABOM is documented in the case's linguistic record. The sequence matters: his analysis preceded Kaczynski's identification by a few months, and his profile was consistent with Kaczynski in general demographic terms. The specific identification, however, came from David Kaczynski, not from any linguistic analysis. Foster later faced criticism in other cases for over-confident attributions, a separate issue that underscores the general principle that stylometric claims need validated error rates.

The UNABOM case is often cited as proof that forensic linguistics works. That framing needs precision. What the case demonstrates is that: (1) stylometric analysis can generate a precise authorial profile from a large text; (2) once a candidate is identified by other means, linguistic comparison of that candidate's known writings to the questioned text can produce strong confirmatory evidence; (3) the linguistic evidence was solid enough that it could withstand scrutiny at multiple levels of analysis by multiple independent analysts.

What the case does not demonstrate: that stylometry can identify an unknown author from the general population without a prior candidate. The FBI had a Manifesto and a profile for seventeen years without producing a name. The name came from a human witness who read the text and recognised specific phrases. Without that tip, the linguistic analysis would have remained a detailed profile of an unknown author. This is the honest operating limit, and it is exactly what critics of overreaching forensic-linguistic claims cite when they demand that attribution reports state their scope explicitly.

First: convergence across independent feature types is more reliable than any single analysis. The UNABOM match was not based on one phrase or one spelling preference. It rested on phrase matches, syntactic patterns, lexical idiosyncrasies, and spelling habits, each independently observed by multiple analysts. When multiple methods point the same direction, the result is less likely to be an artefact of one method's idiosyncrasies.

Second: authorial profiling (generating demographic and psychological inferences from style) is a different enterprise from attribution (naming a specific author), and the two should not be conflated in reports. The FBI's UNABOM profile correctly inferred education level and age range, but could not generate a name. Attribution requires a comparison corpus; profiling does not.

Third: the size and nature of the reference corpus is critical. The comparison to Kaczynski's private journals and academic papers was powerful because those texts were voluminous, span decades, cover multiple genres, and are unambiguously attributed. A comparison to a few emails or a short letter would have been far less conclusive at the same level of confidence.