. Many hypotheses have been developed about the Voynich manuscript's "language", called
Voynichese.
Ciphers According to the "letter-based cipher" theory, the Voynich manuscript contains a meaningful text in some European language that was intentionally rendered obscure by mapping it to the Voynich manuscript "alphabet" through a
cipher of some sort—an
algorithm that operated on individual letters. This was the working hypothesis for most 20th-century deciphering attempts, including an informal team of
NSA cryptographers led by
William F. Friedman in the early 1950s. A 2025 study proposed a historically plausible verbose substitution cipher ("Naibbe") that can encode Latin and Italian as ciphertext, would be possible with 15th century supplies, and exhibits many of the manuscript's known statistical properties, though the author stresses it is a proof of concept and does not claim it is the manuscript's actual cipher.
Shorthand In 1943, Joseph Martin Feely suggested that the manuscript might be a scientific diary, written in a private
shorthand system, using abbreviations, for a language such as Latin. According to
Mary D'Imperio, "other scholars ... unanimously rejected" Feely's proposed readings of the text and the hypothesis that the text was composed using a "system of abbreviated forms" was "not considered acceptable".
Natural language Statistical analysis of the text reveals patterns similar to those of
natural languages. Amancio
et al. (2013) argued that the Voynich manuscript "is mostly compatible with natural languages and incompatible with random texts". The linguist
Jacques Guy once suggested that the Voynich manuscript text could be some little-known natural language, written
plaintext with an invented alphabet. He suggested Chinese in jest, but later comparison of word length statistics with Vietnamese and Chinese led him to take the hypothesis seriously. In many language families of East and Central Asia, mainly
Sino-Tibetan (
Chinese,
Tibetan, and
Burmese),
Austroasiatic (
Vietnamese,
Khmer, etc.) and possibly
Tai (
Thai,
Lao, etc.),
morphemes generally have only one
syllable. Child (1976), a linguist of Indo-European languages for the U.S.
National Security Agency, proposed that the manuscript was written in a "hitherto unknown North Germanic dialect". He identified in the manuscript a "skeletal syntax several elements of which are reminiscent of certain Germanic languages", while the content is expressed using "a great deal of obscurity". In January 2014, Stephen Bax of the
University of Bedfordshire published his research on using a "bottom up" approach to decipher the manuscript. His method involved looking for and translating
proper nouns, in association with relevant illustrations, in the context of other languages of the same time period. A paper he posted online offers tentative translation of 14 characters and 10 words. He suggested the text is a treatise on nature written in a natural language, rather than a code, Tucker & Talbert (2014) published a paper claiming a positive identification of 37 plants, 6 animals, and one mineral referenced in the manuscript to plant drawings in the
Libellus de Medicinalibus Indorum Herbis or Badianus manuscript, an Aztec
herbal written in 1552. Together with the presence of
atacamite in the paint, they argue that the plants were from colonial
New Spain and the text represented
Nahuatl, the language of the
Aztecs. They date the manuscript to between 1521 (the date of the
Spanish conquest of the Aztec Empire) and
circa 1576. These dates contradict the earlier radiocarbon date of the vellum and other elements of the manuscript; however, they argued that the vellum could have been stored and used at a later date. The analysis has been criticised by other Voynich manuscript researchers, who argued that a skilled forger could construct plants that coincidentally have a passing resemblance to theretofore undiscovered existing plants. Nahuatl specialist M. Pharao Hansen has rejected their proposed readings as resting on "pure speculation".
Constructed language The peculiar internal structure of Voynich manuscript words led William F. Friedman to conjecture that the text could be a
constructed language. In 1950, Friedman asked the British army officer
John Tiltman to analyse a few pages of the text, but Tiltman did not share this conclusion. In a paper in 1967, Tiltman said: The concept of a constructed language is quite old, as attested by Wilkins's
Philosophical Language (1668), but still postdates the generally accepted origin of the Voynich manuscript by two centuries. In most known examples, categories are subdivided by adding
suffixes (
fusional languages); as a consequence, a text in a particular subject would have many words with similar prefixes—for example, all plant names would begin with similar letters, and likewise for all diseases, etc. This feature could then explain the repetitive nature of the Voynich text; however, no one has been able yet to assign a plausible meaning to any prefix or suffix in the Voynich manuscript.
Hoax The fact that the manuscript has defied decipherment thus far has led various scholars to propose that the text does not contain meaningful content in the first place, implying that it may be a medieval
hoax. In 2003, computer scientist
Gordon Rugg showed that text with characteristics similar to the Voynich manuscript could have been produced using a table of word prefixes, stems, and suffixes, which would have been selected and combined by means of a perforated paper overlay. The latter device, known as a
Cardan grille, was invented around 1550 as an encryption tool, more than 100 years after the estimated creation date of the Voynich manuscript. Some maintain that the similarity between the pseudo-texts generated in Gordon Rugg's experiments and the Voynich manuscript is superficial, and the grille method could be used to emulate any language to a certain degree. In April 2007, a study by Austrian researcher Andreas Schinner published in
Cryptologia supported the hoax hypothesis. Schinner posited that the statistical properties of the manuscript's text were more consistent with meaningless gibberish produced using a quasi-
stochastic method, such as the one described by Rugg, than with Latin and medieval German texts. Some scholars have claimed that the manuscript's text appears too sophisticated to be a hoax. In 2013, Marcelo Montemurro, a theoretical physicist from the
University of Manchester, published findings claiming that
semantic networks exist in the text of the manuscript, such as content-bearing words occurring in a clustered pattern, or new words being used when there was a shift in topic. With this evidence, he believes it unlikely that these features were intentionally "incorporated" into the text to make a hoax more realistic, as most of the required academic knowledge of these structures did not exist at the time the Voynich manuscript would have been written. In 2021, researchers at
Yale University, using the
tf–idf analysis, further investigated the relation between clusters of subjects in the text and topics as they could be identified by illustrations and
paleography analysis. Their conclusion is that clusters derived by computation match with the topics of the illustrations to some degree, thus providing evidence that the Voynich manuscript contains meaningful text. Other scholars have argued that such sophisticated patterns could also appear in hoaxed documents. In 2016, Gordon Rugg and Gavin Taylor published another article in
Cryptologia demonstrating that the grille method could reproduce many larger-scale features of the text. In 2019, Torsten Timm and Andreas Schinner published a paper arguing that the text was produced by a process of "self-citation" in which scribes copied and modified meaningless words from earlier in the text. Using a computer simulation of this process, they demonstrated that it could reproduce many of the statistical characteristics of the Voynich manuscript. In 2022,
Yale University researchers Daniel Gaskell and Claire Bowern published the results of an experiment in which human participants intentionally tried to write meaningless text. They found that the resulting text was often highly non-random and exhibited many of the same unusual statistical properties as the Voynich manuscript, supporting the idea that some features of the text could have been produced in a hoax.
Glossolalia In their 2004 book, Gerry Kennedy and Rob Churchill suggest the possibility that the Voynich manuscript may be a case of
glossolalia (speaking-in-tongues),
channelling, or
outsider art. If so, the author felt compelled to write large amounts of text in a manner which resembles
stream of consciousness, either because of voices heard or because of an urge. This often takes place in an invented language in glossolalia, usually made up of fragments of the author's own language, although invented scripts for this purpose are rare. Kennedy and Churchill use
Hildegard von Bingen's works to point out similarities between the Voynich manuscript and the illustrations that she drew when she was suffering from severe bouts of
migraine, which can induce a trance-like state prone to glossolalia. Prominent features found in both are abundant "streams of stars", and the repetitive nature of the "
nymphs" in the balneological section. The theory is controversial, and is virtually impossible to prove or disprove, short of deciphering the text. Kennedy and Churchill are themselves not convinced of the hypothesis, but consider it plausible. In the culminating chapter of their work, Kennedy states his belief that it is a hoax or forgery. Churchill acknowledges the possibility that the manuscript is either a synthetic forgotten language (as advanced by Friedman), or else a forgery, as the preeminent theory. He concludes that, if the manuscript is a genuine creation, mental illness or delusion seems to have affected the author. == Decipherment claims ==