Examinations
Forensic authorship analysis: can a linguist show who wrote an email or a text?
Forensic authorship analysis compares the language of a questioned email or message with known writing of each candidate author. Federal Rule of Evidence 901(b)(3) lets an expert make that comparison, and the advisory committee note says language patterns may indicate authenticity or its opposite. Federal courts have admitted linguists' comparisons of features and have limited their conclusions about who the author is.

What questions can forensic authorship analysis answer?
Three kinds: whether two or more texts share an author, which of several known candidates wrote a text, and what the language says about an unknown writer, for example the region the writer comes from.
Dana Roemling and Jack Grieve describe the three tasks in a CREST Security Review article published on March 18, 2024. Verification asks "whether different texts were authored by the same individual." Attribution looks at a set of potential authors and asks which of them wrote the text. Profiling tries "to infer characteristics about the author by their language use."
In litigation the first two tasks match the questions the parties raise: a party denies sending threatening emails, or two anonymous messages are said to come from one person. The linguist compares the questioned texts with known writing and reports which features match, which differ and how common each feature is in comparable writing.
Profiling narrows a list of suspects. Roemling and Grieve cite a ransom note that asked for the money to be left "in the green trash kan on the devil strip at corner of 18th and Carlson." In their words, "devil strip" is "highly regionally bound and primarily used in the city of Akron, Ohio." The misspelled "kan" stood next to hard words spelled correctly, so it may have been a disguise.
How does an authorship comparison work on emails and texts?
The linguist lists features of the questioned messages, such as spelling, punctuation, abbreviations and word choice, and then checks how each candidate uses the same features in writing whose authorship nobody disputes.
Roemling and Grieve describe the Starbuck case, which turned on semicolons. Jamie Starbuck murdered his wife Debbie and then wrote emails in her name. His undisputed emails had few semicolons and hers had many. The questioned emails had even more, but the semicolons "were used grammatically in the same way as Jamie, as opposed to Debbie." He had raised the count to imitate her and kept his own way of using them.
The case is the method in small form. The linguist counted one feature in three groups of texts: the questioned emails, the known emails of the husband and the known emails of the wife. Then the linguist compared where each writer put the semicolon in a sentence.
Short texts limit the method. Roemling and Grieve name sparse data, genre constraints and texts written by several people among its limits. A thread of one-line text messages gives fewer features than a long email, and a formal letter by the same person may not match the way that person writes in a chat.
What does the linguist need from the client?
The questioned messages in the fullest form available, the list of candidate authors, and known writing of each candidate, preferably of the same kind and period, with a record of where each text came from.
Known writing is the base of the comparison. For emails it can be messages a candidate admits sending, messages the candidate produced in discovery, or business correspondence whose authorship nobody contests. For chat messages it can be the candidate's other conversations on the same platform.
The report has to name every text used. Under Rule 26(a)(2)(B), the written report of a retained expert contains "a complete statement of all opinions the witness will express and the basis and reasons for them" and "the facts or data considered by the witness in forming them." A known text that the linguist read and then left out of the comparison still belongs on that list.
The source of each known text matters as much as its content. If the other side disputes that a known email was written by the candidate, the comparison inherits that dispute. Rule 901(b)(3) speaks of a comparison "with an authenticated specimen," so the specimen itself is authenticated first.
How do the rules of evidence treat language patterns?
Rule 901 lets a party authenticate a writing through an expert's comparison with an authenticated specimen and through its distinctive characteristics. The advisory committee note says language patterns may indicate authenticity or its opposite.
Rule 901(a) requires the proponent to "produce evidence sufficient to support a finding that the item is what the proponent claims it is." Rule 901(b) lists examples. Example (3) is "a comparison with an authenticated specimen by an expert witness or the trier of fact." Example (4) covers "the appearance, contents, substance, internal patterns, or other distinctive characteristics of the item, taken together with all the circumstances."
The 1972 advisory committee note to example (4) says: "Language patterns may indicate authenticity or its opposite." It cites Magnuson v. State, 187 Wis. 122, 203 N.W. 749 (1925), and the article by Arens and Meadow, Psycholinguistics and the Confession Dilemma, 56 Colum. L. Rev. 19 (1956). The note to example (3) says it "sets no higher standard for handwriting specimens and treats all comparison situations alike," with Rule 104(b) governing.
Authentication sets a low bar: evidence sufficient to support a finding. An opinion that a particular person wrote a message is expert testimony, so it also has to pass Rule 702. Federal courts have drawn their limits at that second step.
What did the court allow in United States v. Van Wyk?
Judge Bassler let an FBI agent compare markers between known and questioned writings in a threats case. The court barred the agent's conclusion that the defendant wrote the threatening letters.
United States v. Van Wyk, 83 F. Supp. 2d 515 (D.N.J. 2000), is an opinion by Judge Bassler dated February 8, 2000. The defendant was charged with making threatening communications under 18 U.S.C. 875. The government offered FBI Special Agent James R. Fitzgerald, first as an expert in forensic stylistics and then as an expert in text analysis. He examined 22 letters sorted into known typings, known writings, questioned typings and questioned writings.
The court found Fitzgerald qualified in text analysis and left the gaps in his training to cross-examination. On reliability it noted that neither he nor the government identified a known error rate or the number of samples needed to reach a probability of authorship, and that there was "no universally recognized standard for certifying an individual as an expert in forensic stylistics."
The order permitted testimony on "the comparison of characteristics or 'markers' between the handwritten and typed writings" and on how common a given feature was in his experience. It also stated: "Agent Fitzgerald is not permitted to testify to his conclusion as to the identity of the author of the unknown writings." Testimony built on the defendant's history and on the victims was excluded under Rules 403 and 404(b).
What did the court allow in United States v. Zajac?
In 2010 a federal court in Utah excluded the analysis of one letter because of an analytical gap, allowed testimony on similarities between two other letters and barred the opinion that one person wrote both.
United States v. Zajac, 748 F. Supp. 2d 1340 (D. Utah 2010), was decided on September 2, 2010 by Judge Clark Waddoups. Fitzgerald, who had worked at the FBI between 1987 and 2007 and then as a consultant and forensic linguist, compared three letters from Downers Grove, Hinsdale and Salt Lake City and their envelopes. He concluded "with a likelihood bordering on certainty" that one person wrote them.
Zajac contended that Fitzgerald's methods were unreliable because "he failed to conduct blind studies" and "knew of no error rate in the field." Zajac also called William Gregory Eggington, a professor of linguistics at Brigham Young University. The court held that Eggington's attacks on Fitzgerald's methods and bias "go more towards application and weight than invalidation of the protocol as a whole." For the Downers Grove letter the court found "the analytical gap between the data and opinion proffered too great to be admissible" and excluded all testimony about it.
Fitzgerald could testify "about the similarities between the Hinsdale letter and the Salt Lake City letter, but may not offer the ultimate opinion that the two letters were authored by the same person." The court also barred the profiling opinion that the author "is a native English speaker in his or her mid-thirties or older" and any testimony about similarities between the envelopes.
What does Rule 702 require from an authorship opinion?
Under the 2023 text, the party offering the linguist persuades the court that it is more likely than not that the opinion rests on sufficient facts or data and on reliable methods reliably applied to the texts.
Rule 702 covers a witness "qualified as an expert by knowledge, skill, experience, training, or education." Under paragraphs (b) through (d), the testimony must be "based on sufficient facts or data," must be "the product of reliable principles and methods," and the opinion must apply those principles and methods reliably to the facts of the case. The committee note to the 2000 amendment says that in Kumho Tire the Supreme Court clarified that the gatekeeping applies to all expert testimony, including testimony that does not rest on science.
For a linguist who works from experience, the same note says: "If the witness is relying solely or primarily on experience," then the witness "must explain how that experience leads to the conclusion reached." In Van Wyk and in Zajac the court asked how many texts were compared, whether the method had a known error rate and whether the conclusion followed from the features counted.
Both courts admitted the comparison of features and excluded the step that turns those features into a named author or into a single author of two letters. In both cases the order that excluded the conclusion left the feature comparison admissible, so a report for a federal court can set out the comparison and the frequency data separately from the conclusion.
Does authorship analysis replace phone and account records?
No. Rule 901(b)(4) reads distinctive characteristics together with all the circumstances, and account records, device data and witness testimony belong to those circumstances. The linguist examines only the language of the messages.
Rule 901(b)(1) permits authentication by "testimony that an item is what it is claimed to be," for example from the person who received the message. The advisory committee note to example (4) gives two more routes: a document that discloses knowledge only its author would have, and a letter shown to be a reply to an authenticated letter.
Who controlled a phone or an account when a message was sent is a question for a digital forensic examiner and for records of the carrier or the platform. The linguist answers a narrower question: whether the language of the message fits the known writing of a candidate. In the Starbuck case the emails came in the wife's name, and the language pointed to the husband.
What must the report disclose, and how is it answered?
A retained linguist's report states each opinion with its basis, the texts considered, exhibits, qualifications, prior testimony and compensation. A rebuttal report on the same subject is due within 30 days unless the court sets another date.
Rule 26(a)(2)(B) lists six parts: the opinions with "the basis and reasons for them"; "the facts or data considered"; "any exhibits that will be used to summarize or support them"; the witness's qualifications, "including a list of all publications authored in the previous 10 years"; a list of cases in which, "during the previous 4 years, the witness testified as an expert"; and "a statement of the compensation to be paid for the study and testimony in the case."
Under Rule 26(a)(2)(D)(ii), evidence "intended solely to contradict or rebut evidence on the same subject matter" is disclosed "within 30 days after the other party's disclosure," unless the parties stipulate or the court orders otherwise.
A rebuttal linguist works from the same known writings and lists any texts added. The rebuttal can test whether the features called distinctive also appear in other people's writing on the same platform, whether the known texts are of the same genre as the questioned ones and whether the conclusion goes further than the courts allowed in Van Wyk and Zajac.
What does IAE prepare when the authorship of a message is disputed?
An IAE linguist compares questioned emails or messages with known writing of each candidate author and prepares a report on the client's questions or a review of the opposing linguist's report, in up to 5 business days.
The authorship of anonymous messages is one of the subjects on the IAE forensic linguistics page, together with disputed statements, contract wording and trademark similarity. Questioned emails or messages are compared with known writing of each candidate author, and the report lists the known texts used, as Rule 26(a)(2)(B)(ii) requires for facts or data considered. IAE puts questions to the linguist about the language of the text and leaves legal conclusions to the court.
The client sends the questioned messages, the known writings with a note on where each came from and, where there is one, the opposing linguist's report. The secretariat confirms receipt the same day. Within one business day IAE confirms the format, the cost, the deadline and the specialist from its registry, and the client approves the candidate and the CV before work starts. The document is issued in up to 5 business days after the format and cost are confirmed, with a 48-hour rush for tight deadlines.
Every specialist is a member of the public IAE registry, can be verified by Member ID or QR code and confirms in the document that he or she has no interest in the outcome of the case. Materials are not passed to third parties without the client's consent. A report is evidence that the court weighs with the other evidence, and neither IAE nor the linguist guarantees the outcome. IAE is not a law firm and does not advise on case strategy.
Sources
- Roemling D., Grieve J., Forensic authorship analysis, CREST Security Review, March 18, 2024
- Federal Rules of Evidence, Rule 901, Authenticating or identifying evidence (LII)
- Federal Rules of Evidence, Rule 702, Testimony by expert witnesses (LII)
- Federal Rules of Civil Procedure, Rule 26 (LII)
- United States v. Van Wyk, 83 F. Supp. 2d 515 (D.N.J. 2000) (Justia)
- United States v. Zajac, 748 F. Supp. 2d 1340 (D. Utah 2010)
Read also
All articles
Puffery or false advertising: where is the line, and what does a linguist examine in the claim?
"Better Ingredients. Better Pizza." was puffery, and so was "America's Favorite Pasta." "50% Less Mowing" was a statement of fact. The line between
What is a true threat, when is an angry message a crime, and what does a linguist examine in the words?
A true threat is a serious expression of an intent to commit unlawful violence against a particular person or group. The First Amendment does not
Is satire or parody defamation, how do courts read a joke about a real person, and what does a linguist examine?
Satire about a real person is not defamation if a reasonable reader could not take it as stating actual facts. That is the test the D.C. CircuitExpert examinations
Ready to discuss your case?
Send the materials: the secretariat will confirm the format, cost, deadline and the specialist within one business day.

