Within Later Forgeries

What Linguistic Testing Found in MJ 12 Papers

A preliminary forensic linguistics study correctly flagged known frauds and found broader authorship mismatches across several later MJ-12 documents.

177 sources 3 graphics
Preview for What Linguistic Testing Found in MJ 12 Papers

On this page

  • How the authorship comparison was structured
  • Results for known frauds and disputed papers
  • What the study could and could not establish

Introduction

Carole Chaski’s authorship testing is one of the more consequential examinations of the later MJ-12 document corpus because it asked a narrow, testable question: did documents attributed to named historical figures actually resemble authenticated writing by those people? In 2007, Michael S. Heiser published a preliminary report based on computational analysis performed by Chaski, a forensic linguist specialising in authorship attribution. Seventeen disputed documents attributed to nine people were compared with authenticated writing samples. Only one placed its claimed author first in the study’s similarity ranking. More importantly, four documents already identified independently as fraudulent were included without Chaski being told which they were; her test rejected the claimed authorship of all four.[Dr. MSH]drmsh.comDr. MSHMicrosoft WordMJ Test article.doc…

Authorship Tests illustration 1
Explanatory illustration 1

That does not linguistically disprove every alleged MJ-12 document. Several famous papers could not be tested at all, and the exercise was explicitly described as preliminary. What it does provide is an unusual internal check on part of the later corpus: a method capable of recognising known frauds also produced widespread authorship mismatches among documents whose status was still disputed.

How the authorship comparison was structured

The 2007 report, The Majestic Documents: A Forensic Linguistic Report, was written by Heiser on the basis of Chaski’s analysis. The distinction matters: this was not a Chaski-authored journal article announcing a comprehensive scientific verdict on MJ-12. Heiser described the project itself as preliminary because cost prevented the use of the full range of Chaski’s authorship-analysis capabilities. He later continued to describe the paper as his account of Chaski’s analysis.[Dr. MSH]drmsh.comDr. MSHMicrosoft WordMJ Test article.doc…

The basic design nevertheless addressed a real weakness in earlier arguments over the papers. Rather than asking whether a memo looked like a 1940s government document, the test concentrated on its prose. Chaski’s wider work in forensic linguistics stressed empirical authorship comparison rather than intuitive reliance on conspicuous vocabulary, spelling habits or isolated stylistic quirks. The US National Institute of Justice’s summary of her earlier research likewise warns that authorship conclusions should not rest on a single linguistic attribute or common-sense impressions of writing style.[National Institute of Justice]nij.ojp.govNational Institute of JusticeDaubert-Inspired Assessment of Language-Based Author Identification | National Institute of JusticeJanuary 1…

For the MJ-12 exercise, the questioned texts were transcribed into plain-text files. Heiser and his collaborators retained the prose, salutations, misspellings and grammatical errors but removed such non-prose features as stamps, dates, file numbers, handwritten annotations and memorandum headings. That meant the analysis was deliberately separated from questions about typewriters, security markings, signatures, paper or other physical characteristics.[Dr. MSH]drmsh.comDr. MSHMicrosoft WordMJ Test article.doc…

Seventeen questioned documents, attributed to nine claimed authors, qualified for testing. Those names included Franklin D. Roosevelt, Harry Truman, Dwight Eisenhower, John F. Kennedy, George C. Marshall, Roscoe Hillenkoetter, Nathan Twining, Vannevar Bush and Allen Dulles. Thirty authenticated documents by those nine people formed the comparison material. Heiser says the known samples were selected with attention to such factors as genre, period, recipient and text length.[Dr. MSH]drmsh.comDr. MSHMicrosoft WordMJ Test article.doc…

The comparison used an n-gram approach, looking computationally for recurring sequences of linguistic elements and measuring their similarity. Authenticated documents by each historical figure were combined into a known-author pool. Each questioned document was then compared with the pools, and the results were ranked by similarity. Under the validation standard reported for this version of Chaski’s method, the genuine author occupied the most similar, or “top”, position in 90.7 per cent of validation tests. A second- or third-place finish was therefore not treated as partial authentication.[Dr. MSH]drmsh.comDr. MSHMicrosoft WordMJ Test article.doc…

That procedure is particularly relevant to MJ-12 because the corpus contains documents that can imitate the appearance of official correspondence while making specific claims about who wrote them. Stylometry provides a line of inquiry that a convincing letterhead, period typeface or plausible bureaucratic vocabulary cannot by itself satisfy.

Known frauds became an internal check

The strongest feature of the exercise was not simply that most questioned documents failed. Heiser had quietly placed previously exposed fraudulent MJ-12 papers in the test set and says he did not identify them to Chaski beforehand. In effect, these documents served as controls for an obvious question: would the linguistic procedure reject papers already known by independent means to carry false authorship claims?[Dr. MSH]drmsh.comDr. MSHMicrosoft WordMJ Test article.doc…

According to the report, it did. All four such documents failed to match their purported authors: a July 1947 Truman-to-Twining document, the Marshall-to-Humelsine paper of September 1947, a Hillenkoetter-to-Truman document and the so-called Nathan Twining “White Hot Report”. Heiser identified these as documents whose fraudulent nature Stanton Friedman had already detected separately.[Dr. MSH]drmsh.comDr. MSHMicrosoft WordMJ Test article.doc…

This is important because those results were not produced merely by noticing obvious typographical defects in the forged papers. The test material excluded formatting and other visual clues and concentrated on prose. Nor, according to Heiser’s account, was Chaski told which items were the planted known frauds. Their rejection therefore provides a limited but meaningful check that the test was responding to something relevant to authorship rather than simply reproducing Heiser’s prior classifications.

It should not be overstated as a full blind validation experiment. Heiser selected and prepared the corpus, the sample was small, and the MJ-12-specific exercise was preliminary rather than a large independent replication. But within those limits, the known-forgery result is stronger evidence than merely applying a classifier to disputed papers and declaring its output persuasive.

Authorship Tests illustration 2
Explanatory illustration 2

The disputed papers showed a broader mismatch

The result that widened the significance beyond the already exposed frauds was the poor performance of the remaining questioned documents. Of all 17 tested papers, only Twining-Q3, “Twining to General Schulgen,” placed Nathan Twining’s authenticated writing pool in the top position. Heiser consequently reported that this was the only tested item showing a strong likelihood of authorship by the person whose name it bore under the study’s stated criterion.[Dr. MSH]drmsh.comDr. MSHMicrosoft WordMJ Test article.doc…

The other results included striking mismatches. For example, the Roosevelt-attributed FDR-Q1 ranked the Kennedy comparison pool first, followed by Eisenhower and Twining; Roosevelt’s own pool came fourth. A Vannevar Bush-attributed document likewise ranked the Kennedy pool first while Bush’s authenticated material ranked eighth. An Allen Dulles-attributed paper ranked Kennedy first and Dulles fourth. These rankings do not mean that Kennedy literally wrote the papers. They mean that, among the comparison pools available to the test, their measured linguistic characteristics resembled the Kennedy sample more closely than those of their alleged authors.[Dr. MSH]drmsh.comDr. MSHMicrosoft WordMJ Test article.doc…

That distinction becomes especially important because the Kennedy pool ranked first for eight of the 17 questioned documents. Heiser suggested that this pattern might indicate that several purportedly different historical authors were actually being imitated by one common writer. But he explicitly presented that as a suggestion rather than proof.[Dr. MSH]drmsh.comDr. MSHMicrosoft WordMJ Test article.doc…

The cautious interpretation is therefore the stronger one. Eight papers converging on the same comparison pool is suspicious when those papers purport to come from different people, but the experiment was designed principally to test claimed authorship, not to identify an unknown forger. “Closest to Kennedy” is not equivalent to “written by Kennedy”, nor does it independently establish that all eight came from one hand. Demonstrating common anonymous authorship would require a test designed specifically for that question and preferably additional comparison material.

What the study could not test

Some of the best-known alleged MJ-12 papers fall outside the 17-document result. Authorship attribution requires a meaningful claim about who wrote a text, together with enough authenticated writing from that person for comparison. The Eisenhower Briefing Document was excluded because it says it was prepared for Eisenhower rather than identifying Eisenhower as its author. SOM1-01, the alleged special-operations manual on extraterrestrial recovery, similarly supplies no individual author for comparison. The Einstein-Oppenheimer paper was excluded because of its overlapping authorship, while very short documents could also fail the minimum-text requirement.[Dr. MSH]drmsh.comDr. MSHMicrosoft WordMJ Test article.doc…

Selection was not random either. Among documents that had an attributed author and sufficient prose, Heiser gave preference partly to papers making particularly consequential claims about extraterrestrial biological entities, extraterrestrial wreckage or related matters. Cost was another constraint. This makes the exercise better understood as a targeted forensic examination of a significant subset than as a statistical sample representative of the entire MJ-12 archive.[Dr. MSH]drmsh.comDr. MSHMicrosoft WordMJ Test article.doc…

There are methodological reasons for restraint as well. Chaski’s published work treated computational authorship attribution as something that should operate alongside other forensic evidence, not replace it. In her 2007 discussion of authorship identification, she specifically argued for precautions against confirmation bias and for independent convergence between linguistic and digital-forensic findings.[IFIP Digital Library]dl.ifip.orgDigital Library Chapter 9THE KEYBOARD DILEMMA AND AUTHORSHIP IDENNovember 29, 2025…Published: November 29, 2025 The MJ-12 report itself says that only a variation of her fuller ALIAS methodology was affordable and calls the resulting tests preliminary.[Dr. MSH]drmsh.comDr. MSHMicrosoft WordMJ Test article.doc…

Nor does a successful authorship match authenticate what a document says. A genuine letter by a historical official could contain an error, deception or unverified claim; conversely, authorship testing cannot directly determine the age of paper, provenance of a photocopy or authenticity of a security marking. Heiser explicitly drew this boundary between authorship and content.[Scribd]scribd.comPreparation of the Majestic Documents for Testing The Majestic documents tested by Dr.Read more…

Authorship Tests illustration 3
Explanatory illustration 3

What the results change in the MJ-12 debate

Chaski’s tests matter most when read as corpus evidence rather than a universal verdict. The exercise cannot legitimately be reduced to “linguistics proved all MJ-12 documents fake”. It did not examine all of them, and prominent documents such as the Eisenhower Briefing Document and SOM1-01 were outside the method’s reach. Heiser himself left open the possibility that authorless papers might require different forms of testing.[Dr. MSH]drmsh.comDr. MSHMicrosoft WordMJ Test article.doc…

The finding is nevertheless difficult to dismiss as irrelevant to the later corpus. The tested set contained documents supposedly written by numerous senior American officials across different administrations. Yet 16 of 17 failed to place their purported writer first under the study’s criterion; all four independently known fraudulent controls were among those failures; and eight questioned papers unexpectedly clustered nearest the same Kennedy comparison pool.[Dr. MSH]drmsh.comDr. MSHMicrosoft WordMJ Test article.doc…

That pattern changes the evidential burden. A later MJ-12 document cannot gain much credibility merely because its prose sounds official, fits the developing Majestic narrative or resembles other papers in the collection. Similarity within the corpus may itself be contaminated by fabricated material. The present-day Majestic Documents repository still describes a large collection of purported leaked material and foregrounds authentication research, illustrating why distinguishing individual documents rather than treating “MJ-12” as one homogeneous source remains essential.[Majestic Documents]majesticdocuments.comMajestic DocumentsDocuments | Majestic DocumentsOctober 29, 2018…Published: October 29, 2018

The wider official record reinforces the need for that caution without substituting for the linguistic test. The US Government Accountability Office reported that the agencies it consulted found no evidence that submitted MJ-12 material consisted of documents originally created within the executive branch, while relevant presidential-library searches found no corroborating records for the designation.[Government Accountability Office]gao.govGovernment Accountability Office[Comments on Majestic 12 MaterialGovernment Accountability Office[Comments on Majestic 12 Material

Chaski’s contribution is therefore narrower, and more useful, than a sweeping declaration about Majestic Twelve. Her analysis supplied an independent type of test to a portion of the later corpus, correctly rejected documents already exposed as fraudulent, and then found the same basic authorship problem far more widely. It cannot identify who fabricated the mismatching documents, establish that they all share one source, or adjudicate untested papers. What it does show is that, for most of the 17 documents actually examined, the claimed historical authorship did not survive the linguistic comparison.

Amazon book picks

Further Reading

Books and field guides related to What Linguistic Testing Found in MJ 12 Papers. Use these as the next step if you want deeper reading beyond the article.

eBay marketplace picks

Marketplace Samples

Live-tested eBay searches with available results related to this page.

UsingUSA

Selected fromUFO poster oneBay.co.uk.

Endnotes

1. Source: drmsh.com
Title: Dr. MSHMicrosoft Word
Link:https://drmsh.com/michaelsheiser/MJ%20Test%20article.pdf

Source snippet

MJ Test article.doc...

2. Source: drmsh.com
Title: Dr. MSHNew Article on Majestic-12 Documents
Link:https://drmsh.com/article-majestic12-documents/

Source snippet

Dr. Michael HeiserMichael Heiser · About · Curriculum Vitae · FAQ... Carol Chaski. Carol's expertise is authorship... In case the Majes...

3. Source: dl.ifip.org
Title: Digital Library Chapter 9
Link:https://dl.ifip.org/db/conf/ifip11-9/df2007/Chaski07.pdf

Source snippet

THE KEYBOARD DILEMMA AND AUTHORSHIP IDENNovember 29, 2025...

Published: November 29, 2025

4. Source: scribd.com
Link:https://www.scribd.com/document/225277137/MJ-12-Test-Article

Source snippet

Preparation of the Majestic Documents for Testing The Majestic documents tested by Dr.Read more...

5. Source: es.scribd.com
Title: Lunes 22 de Enero Del 2024
Link:https://es.scribd.com/document/743473161/Lunes-22-de-Enero-Del-2024

6. Source: pt.scribd.com
Title: 10 Cuestion Social y Estudios de Trabajo
Link:https://pt.scribd.com/document/331410971/10-Cuestion-Social-y-Estudios-de-Trabajo

7. Source: fr.scribd.com
Title: UFOmania No 76
Link:https://fr.scribd.com/document/730201850/UFOmania-No-76

8. Source: scribd.com
Link:https://www.scribd.com/document/423689839/chaski2012

9. Source: scribd.com
Title: Advances in Digital Forensics III PDF | PDF | Daubert Standard | Expert Witness
Link:https://www.scribd.com/document/385430844/Advances-in-Digital-Forensics-III-pdf

10. Source: scribd.com
Title: Authorship Attribution in Digital Evidence | PDF | Noun | Adjective
Link:https://www.scribd.com/document/813864151/Whos-At-The-Keyboard-Authorship-Attribution-in-Di

11. Source: fr.scribd.com
Title: Ovni OVNIS Enquete Sur Un Secret d Etats Egon Kragel Yves Couprie 2009 297p
Link:https://fr.scribd.com/document/813576215/Ovni-OVNIS-Enquete-Sur-Un-Secret-d-Etats-Egon-Kragel-Yves-Couprie-2009-297p

12. Source: es.scribd.com
Title: Decanter World Wine Awards 2015 pdf
Link:https://es.scribd.com/document/379553476/Decanter-World-Wine-Awards-2015-pdf

13. Source: scribd.com
Link:https://www.scribd.com/document/694319769/MICHELL-12-Determining-the-Authors

14. Source: pt.scribd.com
Link:https://pt.scribd.com/doc/5014360/SOM1-01-Extraterrestrial-Entities-and-Technology-Recovery-and-Disposal-April-1954-part-1
Published: April 1954

15. Source: nij.ojp.gov
Link:https://nij.ojp.gov/library/publications/daubert-inspired-assessment-language-based-author-identification

Source snippet

National Institute of JusticeDaubert-Inspired Assessment of Language-Based Author Identification | National Institute of JusticeJanuary 1...

16. Source: majesticdocuments.com
Link:https://majesticdocuments.com/documents/

Source snippet

Majestic DocumentsDocuments | Majestic DocumentsOctober 29, 2018...

Published: October 29, 2018

17. Source: majesticdocuments.com
Title: Majestic Documents Authentication Articles | Majestic Documents
Link:https://majesticdocuments.com/documents/document-authentication/authentication-articles/

18. Source: gao.gov
Title: Government Accountability Office[Comments on Majestic 12 Material]
Link:https://www.gao.gov/assets/154832.pdf

19. Source: letsrun.com
Link:https://www.letsrun.com/homepages/archive/2020/06/11

20. Source: majesticdocuments.com
Title: Disclaimer | Majestic Documents
Link:https://majesticdocuments.com/documents/disclaimer/

21. Source: majesticdocuments.com
Title: Witnesses Validating Majestic Documents | Majestic Documents
Link:https://majesticdocuments.com/investigations/witnesses/witnesses-validating-majestic-documents/

22. Source: majesticdocuments.com
Title: FB I Investigation | Majestic Documents
Link:https://majesticdocuments.com/investigations/official/fbi-investigation/

23. Source: majesticdocuments.com
Title: documents dated 1948 1959
Link:https://majesticdocuments.com/documents/majestic-documents/documents-dated-1948-1959/

24. Source: majesticdocuments.com
Title: Document Authentication | Majestic Documents
Link:https://majesticdocuments.com/documents/document-authentication/

25. Source: majesticdocuments.com
Title: Books | Majestic Documents
Link:https://majesticdocuments.com/resources/books-products/

26. Source: majesticdocuments.com
Title: Witnesses | Majestic Documents
Link:https://majesticdocuments.com/investigations/witnesses/

27. Source: majesticdocuments.com
Title: Document Sources | Majestic Documents
Link:https://majesticdocuments.com/documents/document-sources/

28. Source: michaelsheiserfoundation.org
Title: Majestic Documents: Forensic Testing and Media Coverage
Link:https://michaelsheiserfoundation.org/articles/the-majestic-documents-and-mj-12-some-resources/

29. Source: sigmod.org
Link:https://sigmod.org/publications/dblp/db/conf/ifip11-9/df2007.html

30. Source: ojp.gov
Link:https://ojp.gov/library/publications/who-wrote-it-steps-toward-science-authorship-identification

31. Source: researchgate.net
Link:https://www.researchgate.net/profile/Carole-Chaski

32. Source: ojp.gov
Link:https://ojp.gov/ncjrs/virtual-library/abstracts/who-wrote-it-steps-toward-science-authorship-identification

33. Source: ojp.gov
Link:https://ojp.gov/ncjrs/virtual-library/abstracts/linguistic-authentication-and-reliability

34. Source: ojp.gov
Link:https://ojp.gov/ncjrs/virtual-library/abstracts/daubert-inspired-assessment-language-based-author-identification

36. Source: drmsh.com
Title: ROSWEL L AND OPERATION PAPERCLIPA
Link:https://drmsh.com/michaelsheiser/UFO.htm

37. Source: drmsh.com
Title: Roswell Mythos
Link:https://drmsh.com/ufo-religions-2/roswell-mythos/

38. Source: drmsh.com
Link:https://drmsh.com/majestic-12-documents-show-on-coast-to-coast-am-tonight/

39. Source: drmsh.com
Title: the majestic documents and mj 12 some resources
Link:https://drmsh.com/the-majestic-documents-and-mj-12-some-resources/

40. Source: drmsh.com
Title: Who’s Afraid of the Big Bad Paranormal Researcher?
Link:https://drmsh.com/afraid-big-bad-paranormal-researcher/

41. Source: specialoperationsmanual.com
Title: SO M1-01
Link:https://www.specialoperationsmanual.com/

42. Source: fringepop321.com
Title: the majestic documents
Link:https://www.fringepop321.com/the-majestic-documents.htm

Additional References

43. Source: youtube.com
Title: Kostas Perifanos
Link:https://www.youtube.com/watch?v=dBqyvpfHy8k

Source snippet

This lecture on Authorship Attribution in Forensic Linguistics provides direct context for understanding computational stylometry and aut...

44. Source: youtube.com
Title: Authorship Attribution in Forensic Linguistics – Forensic Stylistics
Link:https://www.youtube.com/watch?v=CDZ-Lv1JUAc

Source snippet

Forensic Linguistics -- Introduction, Scope, Law & Linguistic Evidence...

45. Source: youtube.com
Title: Your Prose Has A Fingerprint — Stylometry And Forensic Linguistics Explained
Link:https://www.youtube.com/watch?v=_iSetCwmnUE

Source snippet

Kostas Perifanos - Authorship Attribution Using Python...

46. Source: researchgate.net
Link:https://www.researchgate.net/publication/232635884_Compression_and_stylometry_for_author_identification

47. Source: researchgate.net
Link:https://www.researchgate.net/publication/352964106_Historical_sociolinguistics_and_authorship_elucidation_in_medieval_private_written_correspondence_Theoretical_and_methodological_implications_for_and_from_Forensic_Linguistics

48. Source: researchgate.net
Link:https://www.researchgate.net/publication/220542545_Identifying_Authorship_by_Byte-Level_N-Grams_The_Source_Code_Author_Profile_SCAP_Method

49. Source: researchgate.net
Link:https://www.researchgate.net/publication/40498024_Identifying_reliable_valid_markers_of_authorship_A_response_to_Chaski

50. Source: researchgate.net
Link:https://www.researchgate.net/publication/2588226_Multi-Topic_E-mail_Authorship_Attribution_Forensics

51. Source: researchgate.net
Link:https://www.researchgate.net/publication/299537913_Whose_tweet_authorship_analysis_of_micro-blogs_and_other_short_form_messages

52. Source: researchgate.net
Link:https://www.researchgate.net/publication/295403324_Homotopy_Based_Classification_for_Author_Verification_Task_Notebook_for_PAN_at_CLEF_2015