Successful decipherments share a short list of ingredients: a large enough body of text, a known or related language, and ideally a bilingual inscription. When one or more is missing, even brilliant analysis stalls. That is the lens for this file. We begin with two scripts that were read, then turn to those that have not been.
Two benchmarks: hieroglyphs and Linear B
Solved The Rosetta Stone, found in 1799 near Rashid in the Nile Delta and held in the British Museum since 1802, carries one decree in hieroglyphic, Demotic and Greek. The Greek could be read; it supplied names such as Ptolemy, which gave Thomas Young and then Jean-François Champollion their foothold. Champollion's 1822 Lettre à M. Dacier established that hieroglyphs combined sound signs and meaning signs, and his knowledge of Coptic, a late form of Egyptian, supplied the language.
Source: British Museum, EA 24; J.-F. Champollion, Lettre à M. Dacier (1822).
Linear B had no bilingual. Arthur Evans excavated tablets at Knossos from 1900, and Carl Blegen found more at Pylos in 1939. Alice Kober showed in the 1940s that words changed their endings in regular patterns, the so-called "Kober triplets". Michael Ventris built a grid of sound values on that work and announced in 1952 that the language was an early form of Greek. With John Chadwick he published Documents in Mycenaean Greek in 1956. Later finds confirmed the readings.
Source: Ventris & Chadwick, Documents in Mycenaean Greek (1956); M. Fox, The Riddle of the Labyrinth (2013).
Linear B fell because there was enough text to find patterns and a known language waiting at the end of them. Remove either and the method has nothing to grip.
Riddle Garden Frenzy editorial summary of Chadwick's account in The Decipherment of Linear B (1958)Linear A
Open Linear A, used on Minoan Crete roughly from the 18th to the 15th century BCE, is the ancestor of Linear B. Because many signs look alike, scholars can apply Linear B sound values and pronounce Linear A words with reasonable confidence. The trouble is that the resulting words do not match Greek or any other known language. The corpus is small, around 1,400 inscriptions, most of them short administrative records. Reading sounds is not the same as understanding meaning.
Source: J. Younger, Linear A texts in phonetic transcription (University of Kansas); GORILA corpus, Godart & Olivier.
The Indus script
Open Several thousand inscribed objects survive from the Indus Valley civilisation (c. 2600–1900 BCE), mostly small stamp seals from sites such as Harappa and Mohenjo-daro. Inscriptions average about five signs. There is no bilingual, and the underlying language family is unknown. Iravatham Mahadevan's 1977 concordance and Asko Parpola's corpus volumes made systematic study possible.
The debate here even includes whether the signs are writing. In 2004 Steve Farmer, Richard Sproat and Michael Witzel argued that they were a non-linguistic symbol system. In 2009 Rajesh Rao and colleagues published an entropy analysis in Science suggesting language-like structure; critics replied that the statistics could not distinguish language from other ordered systems. This remains a contested question, not a settled one.
Source: Farmer, Sproat & Witzel, EJVS (2004); Rao et al., Science 324 (2009).
Rongorongo
Open About two dozen wooden objects carved with rongorongo glyphs survive from Rapa Nui (Easter Island), scattered across museums including the Santiago Staff in Chile. Lines run in reverse boustrophedon: the reader turns the tablet upside down at each line. Thomas Barthel identified a lunar calendar sequence on the Mamari tablet, refined by Jacques Guy in 1990. Beyond that, readings remain unproven, and it is still debated whether rongorongo predates European contact in 1722.
The Voynich manuscript
Open Beinecke Rare Book and Manuscript Library MS 408 at Yale University is named after the bookseller Wilfrid Voynich, who acquired it in 1912. Radiocarbon tests at the University of Arizona in 2009 dated its vellum to 1404–1438. That dates the animal skin, not necessarily the writing, although ink analysis has found nothing inconsistent with the period.
The claim
The Voynich manuscript has been "solved" (a claim announced repeatedly, in many languages).
The record
No proposed solution has produced a consistent, repeatable translation that independent specialists accept. Statistical studies show word-frequency patterns resembling natural language, which also makes a meaningless hoax harder, though not impossible.
Under evaluation
In 2022 François Desset and colleagues published a proposed decipherment of Linear Elamite in Zeitschrift für Assyriologie, using inscribed silver beakers with parallel cuneiform names. It has been well received by some specialists and questioned by others. We list it as Contested until independent work confirms or rejects it.
Next, see how surviving texts guide, and sometimes mislead, searches for lost places.