Washington · 21 November 1864 · the dispute over its author from 1904 to 2019
Who composed the 139-word letter of condolence sent to Lydia Bixby of Boston over Abraham Lincoln's signature on 21 November 1864: Lincoln himself, or his private secretary John Hay?
What happened. In September 1864 a Boston widow, Lydia Bixby, showed five letters to William Schouler, the Adjutant General of Massachusetts, the officer who kept the state's military records; each reported the death of one of her sons in the Union army, and Governor John A. Andrew of Massachusetts asked the War Department for a letter from the President. On 21 November a letter of 139 words signed "A. Lincoln" was sent to her, and on 25 November, the day it reached her, the Boston Evening Transcript printed it. It has been reprinted and anthologised ever since, and it was read aloud in the film Saving Private Ryan (1998). Its facts were wrong, since only two of the five sons had been killed, and the error was Schouler's.
Why it is a mystery. The manuscript has never been found: no draft or copy survives, the newspaper text is the only authority for the wording, and the "facsimiles" of the manuscript sold from 1891 onward are forgeries. Lincoln signed the letter, but his 26-year-old private secretary John Hay (later Secretary of State, 1898 to 1905) drafted many letters for Lincoln's signature, and after Hay's death several men said he had told them in confidence that this one was his; none wrote it down in his lifetime, the earliest record is from 1922, and Hay himself wrote in 1904 that "the letter of Mr. Lincoln to Mrs. Bixby is genuine" and told Robert Lincoln, the President's eldest son, that he knew nothing of the letter's source. For Hay: drafting was his job, "beguile", the letter's rarest word, is his and never Lincoln's, and a clipping of the letter is pasted in two of his scrapbooks; the case was built by Michael Burlingame, author of the fullest modern life of Lincoln (1995, 1999), and seemed settled in 2019 when Jack Grieve, a linguist at the University of Birmingham, and six colleagues applied n-gram tracing, a technique that counts how many of a text's short strings of letters and words occur in each candidate's known writing, and found Hay "almost certainly" the author. For Lincoln: he wrote his own condolences (both that survive from his presidency are in his hand), the request came through the Secretary of War with a governor's wish that the President write, and Hay's own statements point away from Hay; the case was argued by F. Lauriston Bullard, a Boston editor (1946), by Joe Nickell, an investigator of questioned documents, with the linguist Jean Pival (1989), and by Jason Emerson, a historian of the Lincoln family (2006, 2008), while Roy P. Basler, editor of the standard edition of Lincoln's writings (1953), printed the letter as Lincoln's and called the claim for Hay "unproved".
How it was judged. The printed record was read in full, every reported claim was dated and traced from mouth to print, and the documents of the week were counted from the notes in Basler's edition (his source codes) that record where each surviving item comes from and whose hand it is in. The 2019 technique was rebuilt from the paper and checked against the paper's results, which reproduce, then tested in three ways the paper did not (the date of the comparison texts, their register, meaning the kind of writing, and manuscript against print), with a second implementation of the tests (the same tests programmed again from the paper, without sight of the first implementation's code) as a check; the technique and three other standard methods were then tried on the known letters Hay wrote for Lincoln's signature, and none detects them. With the stylometry, the statistical comparison of writing habits, treated as neutral, the documentary items were weighed as explicit likelihood ratios, each a stated weight for how far the item favours one man over the other, with a sensitivity table showing what changing any one weight does, and the result is a lean: Lincoln composed the letter, about 53 percent; Hay, about 42; another hand in the office, about 5. The question stays unsettled, since changing any one weight moves Lincoln between 26 and 75 percent; a draft in Hay's hand, or a rerun of the register test on the paper's own corpus (the collection of known texts it compared the letter with) that finds no register effect, would move it toward Hay, and the manuscript in Lincoln's hand with corrections would move it toward Lincoln.
The question is unsettled; on a judgement of the documentary record Lincoln composed the letter, about 53 percent; John Hay, about 42; another hand in the office, about 5; Lincoln against Hay alone, 56 to 44.
The question cannot be settled on the present record: the lean to Lincoln rests on documentary items weighed as likelihood ratios, any one of which moves him between 26 and 75 percent, and no stylometric method run here separates the two men in this register. The added tests exist in two implementations that agree on the direction and differ on the rate; no one else has rerun them, and neither used the paper's own corpus.
Every percentage is conditional. It comes from the documentary items weighed as explicit likelihood ratios (report, section 5), and the weights are judgements. Changing any one weight moves Lincoln between 26 and 75 percent; taking the 2019 stylometry at face value puts him at 11. The sensitivity table is printed beside the verdict. The report leaves Hay possible and does not establish that Lincoln wrote the letter.
The strongest case for Hay is the 2019 n-gram study, which called him "almost certainly its primary author". It reproduces: under its own design (Lincoln before 18 May 1860 against all of Hay) the letter goes to Hay at all seventeen string lengths (the seventeen lengths of letter and word sequence the method counts), and every one of 1,011 known texts of 50 to 400 words is attributed correctly by the majority vote (the majority of the analyses at character-string lengths 4 to 10). Under three tests the paper did not run, its attribution proves sensitive to period, register and edition. Lincoln's own-hand wartime texts of 100 to 200 words go to Hay 14 percent of the time under a pre-1860 Lincoln pool (a pool is the set of a man's known texts that the letter is compared with); 30 of 63 register-matched Lincoln pieces (pieces of his own elevated wartime prose cut to the letter's length, about 140 words each, from 30 documents of 1861 to 1865, the Gettysburg Address among them) go to Hay under that design in the first implementation and 19 of 63 in the second; and 95 Lincoln texts present in both Basler's edition (from manuscripts) and Arthur Brooks Lapsley's 1905 edition (from print) sit 0.010 to 0.016 nearer Hay in print at every string length. The gap between 19 and 30 lies in the Lincoln pool rather than in the code: the one difference that matters is 19 items of Lincoln's 1858 campaign speeches and debates, 83,000 words, which dilute Lincoln's sample (the random draw the method takes from each pool) with oratory and, under the method's rule of comparing equal-sized samples, enlarge Hay's sample by a matching amount, from 159,000 to 243,000 words. With those speeches restored to the second implementation's pool, its own code sends 27 of 63 to Hay. The range, 19 to 30 of 63 (30 to 48 percent), is what the evidence supports.
The methods also cannot detect Hay writing as the President. Under the own-hand design (Lincoln's autograph writing, meaning what survives in his own hand, of 1860 to 1865 against Hay's letters and diary of the same years) the n-gram vote calls three of Hay's four known letters for Lincoln's signature and Hay's own condolence of 1864 Lincoln, and the calibrated classifier (a statistical classifier trained on short character strings, its probabilities adjusted so that a stated 60 percent is right about 60 percent of the time) and Burrows's Delta (a distance measure based on the most common words) call all five Lincoln; the second implementation finds the same under pools of its own matched in date. A result on the letter under these methods is therefore not evidence for either man, and the stylometric likelihood ratio is taken as 1 (range 1:3 to 3:1). Under the own-hand design the letter itself splits eight string lengths to Hay and nine to Lincoln, though by the paper's own rule, the majority of the votes at character-string lengths of 4 to 10, it still goes to Hay, six to one.
The shares come from the documentary items weighed as likelihood ratios in section 5 of the report, starting from 65 percent for Lincoln on the office's practice. The range beside Lincoln's figure is what changing any one weight does; taking the 2019 stylometry at face value would put him at 11. Every figure depends on those weights, which are judgements.
The printed record was read in full: the letters and diary of John Hay, Lincoln's private secretary, printed in 1908; the memoirs of the men who reported that Hay had claimed the letter; the books and articles on both sides, from F. Lauriston Bullard, a Boston editor, in 1946 to Michael Burlingame, author of the fullest modern life of Lincoln, in 1995 and 1999; and Roy P. Basler's standard edition of Lincoln's writings, which records whose hand each surviving document is in. Burlingame's chapter of 2000, the fullest case for Hay, was not available, and four of Basler's volumes could be had only as online lending copies, which cannot be read as full text. No archive was visited.
Every reported claim was dated and traced from mouth to print, and the documents of the week the letter was sent were counted from Basler's notes on the hand each is in. The 2019 study by Jack Grieve, a linguist at the University of Birmingham, and six colleagues had applied n-gram tracing, a technique that counts how many of a text's short strings of letters and words occur in each candidate's known writing, and found Hay the author. That technique was rebuilt from the paper and checked against the paper's results, which reproduce.
It was then tested in three ways the paper did not, with Lincoln texts closer to the letter in date, in kind of writing (its register) and in manuscript against print, and a second implementation of the tests, programmed again from the paper without sight of the first implementation's code, was run as a check; the paper's own corpus could not be obtained. The technique and three other standard methods were then tried on the known letters Hay wrote for Lincoln's signature, and none detects them, so the stylometry, the statistical comparison of writing habits, was treated as neutral.
Figure 1 plots the letter's margin at each of the seventeen string lengths, Lincoln's share of the letter's n-grams minus Hay's, under the 2019 paper's design and under the own-hand design, so that a reader can see how far the attribution to Hay depends on the choice of comparison texts. Figure 2 shows how often the method attributes a known text to the right man at each string length, for Lincoln's ordinary texts and for the elevated wartime pieces of each man, under both designs. The documentary items were weighed as likelihood ratios, each a stated weight for how far the item favours one man over the other; taking the 2019 stylometry at face value would put Lincoln at 11 percent, and discounting every reported claim to nothing puts him at 72.
Under the added tests the 2019 design sends between 19 and 30 of 63 pieces of Lincoln's own wartime prose to Hay, the Gettysburg Address among them, so that study does not settle the question. What remains is documentary and close to a balance: on Hay's side the reported claims and the scrapbook clippings, on Lincoln's the office's practice and Hay's own written statements. Weighing those items gives Lincoln about 53 percent, Hay about 42 and another hand in the office about 5, conditional on the weights, which are judgements; changing any one weight moves Lincoln between 26 and 75 percent, so the question stays unsettled.
Already established by others. The hearsay character of every reported Hay claim, five chains (each a story passed from Hay through one or more people before it was written down), none first-hand, none written down before 1922 (Bullard 1946; Emerson 2008); Hay's letter of 19 January 1904 to William E. Chandler, a former senator from New Hampshire, and what he told Robert Lincoln (Bullard 1946; Emerson 2008); Hay's practice of writing routine letters for Lincoln's signature, four of them in the Bixby week (Basler); the reported claims assembled, "beguile" and the scrapbooks (Burlingame 1995, 1999); the parallels in Lincoln's own letters and the grammatical habits that favour Lincoln, such as his "should" where Hay would write "would" (Nickell and Pival 1989); that the 2019 result reproduces under its own design and passes its own validation (Grieve and colleagues 2019); that n-gram tracing is "highly sensitive to the choice of authors and texts in the comparison corpus" (an earlier test of the method on literary texts by Thomas Proisl and colleagues, computational linguists in Germany, 2018); and that the method's sensitivity to register needed testing, a concern the authors raised themselves at Aston University on 13 April 2016, so that what is new here is the specific test rather than the concern.
What historians have said. That authorship is unproved: Basler's 1953 headnote (the note heading the letter in his edition) called the claim for Hay "unproved"; Jason Emerson (American Heritage 2006; Lincoln Herald 2008) and Harold Holzer (a review on the history website Emerging Civil War, 23 July 2026) still call the question unsettled. Within that unsettled question the balance had tilted to Hay since 2019, on Michael Burlingame's articles ("New Light on the Bixby Letter", Journal of the Abraham Lincoln Association 1995; "The Trouble With the Bixby Letter", American Heritage 1999) and on the n-gram study by Grieve, Clarke, Chiang, Gideon, Heini, Nini and Waibel ("Attributing the Bixby Letter using n-gram tracing", in the journal Digital Scholarship in the Humanities 34, 2019), which called Hay "almost certainly its primary author"; Tyler Dennett, Hay's biographer (1940), and James G. Randall and Richard N. Current, in the last volume of Randall's life of Lincoln (1955), had leaned to Hay earlier. On the other side stood F. Lauriston Bullard (Abraham Lincoln and the Widow Bixby, 1946), Joe Nickell with Jean Pival (Lincoln Herald 1989) and Emerson.
The 2019 attribution reproduces, but its validation does not carry over to texts of the letter's period and register: under the paper's design between 19 and 30 of 63 register-matched Lincoln pieces go to Hay, the Gettysburg Address among them, and the design cannot detect Hay writing as the President. Two implementations of these tests, written independently of each other, agree on the direction (19 of 63 and 30 of 63 on the same pieces; on 97 pieces the second implementation chose itself, 27 by its own code and 28 to 40 by the first's, depending on the pool), and the gap between 19 and 30 is traced to the choice of Lincoln texts: with Lincoln's 1858 campaign speeches restored to its pool, the second implementation's own code sends 27 of 63 to Hay. With the 2019 study treated as neutral, what remains is documentary and near a balance, leaning to Lincoln on judgement weights. It stays provisional because both implementations reconstruct the paper's design on corpora the paper did not use (its own corpus could not be obtained), because the size of the register effect depends on the Lincoln pool, and because no one else has rerun the tests. A rerun on the paper's own corpus by a third party would settle the rate; if it found no register effect, the balance would return to the 2019 tilt to Hay.