In 1803 or 1804, men of the Onslow Company digging the Money Pit lifted out a flat stone cut with rows of strange characters. J. B. McCully, secretary of the Oak Island Association, gave the earliest description in print in 1862: "at 80 feet was a stone cut square, two feet long and about a foot thick, with several characters cut on it." Later accounts put it at ninety feet, and the 90-foot stone is the name that stuck. The stone has been missing since 1919. What survives is a set of symbols that only appeared in writing in the twentieth century, and a famous reading of them: "Forty feet below two million pounds are buried."

This decoder lets you test that reading, and its rivals, for yourself. Choose a recorded version of the symbols and a way of reading them, and your browser searches for the most convincing message in English, French, Dutch, Spanish and Latin as they were written between 1200 and 1900. Each result is then compared with the same symbols in random order, so a reading only counts when it does better than chance.

From the Money Pit to a bookbinder's window

The stone was never sent to a museum. John Smith, one of the three young men who found the Money Pit and later owned the land around it, built it into the fireplace of his house, where visitors came to see it. An anonymous article in the Halifax British Colonist of 2 January 1864 says it could still be seen there "by the curious at the present day." Charles Driscoll wrote that it then spent some time at J. B. McCully's home in Truro.

Around 1865, A. O. Creighton, treasurer of the Oak Island Association, took it to Halifax and set it in the window of A. and H. Creighton, the bookbindery he ran with his brother Herbert, while the association was selling shares. Jefferson W. MacDonald, who helped take it out of the fireplace, told the treasure hunter Frederick Blair in 1894 that "the inscription was easily traced but that no person present could decipher it."

The stone stayed at the bindery for more than fifty years. Harry Marshall, whose father Edward Marshall became a partner in the firm, started work there in 1890. In 1935 he gave Blair and the lawyer R. V. Harris a statement describing a block about two feet long, fifteen inches wide and ten inches thick, weighing about 175 pounds, dark, very hard and with an olive tinge, unlike any local stone. Someone had cut the initials "J.M." on one corner, "but apart from this there was no evidence of any inscription either cut or painted on the stone. It had completely faded out. We used the stone for a beating stone and weight." When the business merged in 1919 the stone was left behind. Searches in 1933 and 1935, including the stone yards of the Brookfield Construction Company in Halifax, found nothing.

Where the symbols come from

No photograph or rubbing of the stone is known. The oldest surviving copy of the symbols is a typescript by the Reverend A. T. Kempton of Cambridge, Massachusetts, dated 28 April 1949 and now among the R. V. Harris Papers at the Nova Scotia Archives. Kempton said he took them from the manuscript of an unnamed schoolteacher. Most readers know the symbols from the drawing in Edward Rowe Snow's True Tales of Buried Treasure, which, according to D'Arcy O'Connor, differs slightly from the manuscript. Most later books copied Snow's drawing.

The copies agree on the message but differ in details that matter to a cipher: a crossed-out triangle at the very start, the shape of the M, and, in the copy Ross Wilhelm of the University of Michigan worked from in 1971, a T and a plus sign where the others show a dagger. The decoder keeps each recorded version separate, so you can test them one against another.

How the stone has been read

The first reading on record appears in a prospectus of the Oak Island Treasure Company in the 1890s, which reported that an expert had read the stone as ten feet below, two million pounds buried. Later writers name him as James Liechti, a professor of languages at Dalhousie College in Halifax, although no document confirms it. Randall Sullivan observes that a reading produced for a company selling shares invites suspicion.

The version that stuck came through Snow. Kempton told him that "an old Irish schoolmaster" had read the symbols as "Forty feet below two million pounds are buried": the same message, with forty feet in place of ten. It is a simple substitution cipher, one symbol for each letter, of the kind found in newspaper puzzles. Paul V. S. Townsend, the cryptographer who worked with Lionel and Patricia Fanthorpe, reached the same reading in under ten minutes, treating the crossed-out triangle as a slip and ignoring it. Joy A. Steele reads the triangle and its crossed-out twin together as a single old-style capital F.

Other readers have found other languages. In 1971 Wilhelm read the stone as Spanish, enciphered with a turning disk of the kind described by the Italian scholar Giambattista della Porta: "A ochenta gui(a) mij(o) r(i)a sumideq(o). F." He took it as directions to guide maize or millet into a drain at eighty feet, with F standing for Philip II, and called his own result "at best a probability statement." In the late 1970s Barry Fell, formerly of Harvard University, received a copy of the symbols from the Nova Scotia Archives. According to Mark Finnan, Fell read them as a religious text in a script used by Coptic Christians, a warning that the people must remember their God or perish.

Every one of these readings depends on the symbols being what the diggers of 1803 saw. The Fanthorpes raise the possibility that someone raising money in 1865 added a message in an easy cipher, and Marshall's statement shows that by the 1890s nothing could be read on the stone at all.

How the decoder reads a message

The decoder judges a reading by how readable it is: how much it looks like real writing from a given language and period. For each of five languages and four periods it carries a model built from texts of the time: how often every group of four letters occurs, and a list of the 20,000 most common words. The texts range from the romances of Chretien de Troyes, the histories and law codes of Alfonso X of Castile, the Ormulum and the works of Thomas Aquinas to the chronicles of Jean Froissart, the 1637 Dutch Bible, the plays of Pedro Calderon de la Barca and nineteenth-century novels. The earliest period, 1200 to 1300, also takes in a few texts from the later twelfth century. The alphabet folds j into i and v into u, as scribes and printers did for most of that time.

A search tries thousands of keys, keeps the readings that look most like the chosen language, and starts again from scratch many times to see whether it keeps arriving at the same answer.

Under each reading, Word meanings looks up every word in Wiktionary among the commonest words of that language and period. It is a dictionary lookup, not a translation: it shows how much of a reading is made of real words, and what those words mean on their own.

Shuffled versions: the test against chance

Forty symbols make a very short text, and a determined search can make almost any short string look like words. So every test runs twice with exactly the same effort: once on the inscription, and then on shuffled versions, the same symbols in random order. A normal search tries ten shuffled versions; a longer one tries twenty. Each result puts the inscription's most readable result next to the most readable result from its shuffled versions. When the inscription is far more readable, the result is marked Meaningful. A small lead is marked Possible, since it can still be chance. When a shuffled version reads as well, the result is marked Chance.

The more freedom a method allows, the more readable the shuffled versions become. Shifting alphabets and cipher disks can turn nearly any symbols into something readable, so their results mean something only when the inscription itself pulls ahead.

What the tests show so far

With the recorded word breaks and one letter for each symbol, the search arrives at the familiar English message without being told what to look for: FORTH FEET BELOW TWO MILLION POUNDS ARE BURIED. The one difference from the traditional reading is the arrow. It appears only once on the stone, so nothing in the inscription decides whether it stands for Y or H. In English of 1500 to 1900 the result is Meaningful, and in English of 1300 to 1500 only Possible. French, Dutch, Spanish and Latin give Chance, and so does every language as written between 1200 and 1300: in thirteenth-century spelling, shuffled versions read as well as the inscription.

Without the word breaks, no language beats chance, English included, so the division into words carries most of what the inscription says. Shifting alphabets and rearranged orders add nothing beyond the traditional reading.

Wilhelm's disk, rebuilt from his own worked tables, gives his Spanish reading back letter for letter, but only from the copy he corrected in two places to make it fit. Under his rules, every language gives Chance.

None of this settles what was cut into the stone. A meaningful English result fits two accounts equally well: an English message in a simple cipher on the stone itself, or a set of symbols drawn up later by someone writing in English. The decoder can show which readings rise above chance. Which account is true is a question for the archives.

Sources

The stone, its history and its readings

J. B. McCully, Halifax Sun and Advisor, 2 July 1862, and "Digging for Buried Treasure at Oak Island, Chester," Yarmouth Herald, 10 July 1862
Anonymous, Halifax British Colonist, 2 January 1864
Harry Marshall, statement to Frederick L. Blair and R. V. Harris, 1935, as printed by William S. Crooker and Randall Sullivan
A. T. Kempton, typescript story of Oak Island, 28 April 1949, Nova Scotia Archives, R. V. Harris Papers MG 1 vol. 384, item 2364f-h
R. V. Harris, The Oak Island Mystery (1958; revised 1967)
Edward Rowe Snow, True Tales of Buried Treasure
Ross Wilhelm, The Spanish in Nova Scotia in the XVI Century: A Hint in the Oak Island Treasure Mystery, University of Michigan Bureau of Business Research Working Paper 23 (1971)
D'Arcy O'Connor, The Money Pit
William S. Crooker, Oak Island Gold (1993)
Lionel and Patricia Fanthorpe, The Oak Island Mystery (1995)
Mark Finnan, Oak Island Secrets
Joy A. Steele and Gordon Fader, Oak Island Mystery Solved, 2nd edition (2018)
Randall Sullivan, The Curse of Oak Island: The Story of the World's Longest Treasure Hunt (2018)

Period texts behind the language models

Distant Reading for European Literary History (COST Action CA16204), European Literary Text Collection: English, French and Spanish novels
Drama Corpora Project (DraCor): English drama from EarlyPrint, Shakespeare, French drama, Calderon and Neo-Latin drama
Simon Gabay and others, FreEM corpora of early modern French
University of Michigan, Corpus of Middle English Prose and Verse
The Latin Library, through the Classical Language Toolkit
Instituut voor de Nederlandse Taal, Corpus Gysseling
Hispanic Seminary of Medieval Studies, Old Spanish Textual Archive (transcriptions by the HSMS, edited by Francisco Gago Jover and F. Javier Pueyo Mena)
Open Medieval French, texts of the twelfth and thirteenth centuries
Index Thomisticus Treebank (Universal Dependencies), the works of Thomas Aquinas
Middle Dutch Sister Books of the Modern Devotion
The Statenvertaling (1637)
Project Gutenberg editions of Juan Ruiz, Don Juan Manuel, Fernando de Rojas, Miguel de Cervantes, Jean Froissart, Christine de Pizan, Francois Villon, Joost van den Vondel, P. C. Hooft, Pieter Langendijk, Betje Wolff and Aagje Deken, Nicolaas Beets, Hendrik Conscience and Multatuli

Word meanings

Wiktionary entries for Latin, Old, Middle and modern French, Middle and modern Dutch, Old and modern Spanish and Middle English, through the kaikki.org extracts made with Tatu Ylonen's wiktextract (CC BY-SA)