Lost Languages: How Linguists Crack Ancient Scripts

Linear B Was Cracked by an Amateur Architect in 1952

On June 24, 1952, Michael Ventris — a 29-year-old architect with no university linguistics degree — broadcast a BBC radio program announcing that Linear B, the script used in Bronze Age Mycenaean Greece, was an early form of Greek. He had spent a decade working on the problem as a side project, using methods borrowed from wartime cryptography. The discovery rewrote the timeline of Greek civilization: it established that literate Greek-speaking palace administrators existed at Knossos and Pylos at least three centuries before Homer. Ventris died in a car accident four years later at 34, before receiving any major academic honor.

His success followed a recognizable decipherment logic. Understanding why it worked — and why other scripts have not yielded to the same approach — illuminates both the power and limits of linguistic detective work.

The Prerequisites of Successful Decipherment

Every successful decipherment of an ancient script has depended on meeting at least one of three conditions: a bilingual or multilingual key text, an identifiable known language encoded in the unknown script, or sufficient repetitive text to reveal statistical structure. Without at least one, decipherment becomes speculation.

• Bilingual/multilingual key: The Rosetta Stone (196 BCE) bore the same priestly decree in Egyptian hieroglyphics, Demotic script, and Greek. Jean-François Champollion used knowledge of Coptic (a descendant of ancient Egyptian) and the Greek text to crack hieroglyphics in 1822
• Known language in unknown script: Linear B — Ventris correctly guessed the underlying language was Greek, letting known Greek vocabulary guide phonetic assignments
• Statistical structure: frequency analysis of symbols, positional patterns, and word boundaries can reveal syllabic vs. alphabetic structure even without translation

The Mayan Decipherment: A Century-Long Project

Mayan glyphs were considered pictographic and undecipherable well into the 20th century, a misconception partly perpetuated by Bishop Diego de Landa, who in 1566 burned hundreds of Maya manuscripts as "works of the devil" while simultaneously — ironically — documenting what he thought was a Mayan "alphabet" (actually a phonetic syllabary misunderstood as alphabetic).

Soviet linguist Yuri Knorozov's 1952 paper demonstrated that Mayan writing was phonetic — specifically syllabic — using Landa's own documentation. Working from photographs in Moscow without ever visiting a Maya site, he correctly identified syllabic readings for several signs. His work was initially rejected by Western Mayanists (partly from Cold War politics) but was eventually vindicated by Tatiana Proskouriakoff's 1960 discovery that Maya inscriptions recorded dynastic history, not astronomical or religious abstraction. By the 1990s, the majority of Maya inscriptions could be read.

Etruscan: The Frustrating Near-Miss

Etruscan is one of the most tantalizing cases in decipherment history. The script is fully readable — it uses a variant of the Greek alphabet, and the phonetic values of all signs are known. The language can be phonetically transcribed. The problem is comprehension: Etruscan is a language isolate with no demonstrated relatives. It is not Indo-European, not Semitic, and not related to any other known language family.

Approximately 13,000 Etruscan inscriptions survive, but 90% are very short funerary texts — names, epithets, dedicatory formulas — that provide insufficient context for semantic analysis. The Pyrgi Tablets (500 BCE), discovered in 1964, contain parallel Etruscan and Phoenician texts. The texts are not exact translations, but they confirmed existing interpretations of some terms. Scholars can read Etruscan as "Vel Partunu was a haruspex of great renown" but they cannot confidently decode most vocabulary beyond names, numbers, and a core of ~250 understood words. Without a long bilingual text in a known language, full comprehension remains out of reach.

Rongorongo: Easter Island's Unsolved Script

Rongorongo — a system of glyphs found on 24–26 surviving wooden tablets and objects from Easter Island (Rapa Nui) — was first brought to the attention of the outside world by Bishop Tepano Jaussen in 1869. By that time, the Peruvian slave raids of 1862–1863 had killed or deported most of the island's population, including virtually everyone who could read the script. The chain of oral transmission was severed within a generation of the script's first Western documentation.

The tablets contain approximately 14,000 glyph tokens using around 120 distinct signs. The script may be proto-writing (recording counts or names) or a full phonetic system — arguments exist for both. No bilingual key, no known underlying language (Rapanui has multiple language influences), and the severely limited corpus make standard decipherment methodology inapplicable. A 2022 study by Ulla Stein and Søren Wichmann applied computational phylogenetic analysis to the glyph ordering and found statistical regularities suggesting genuine linguistic encoding rather than mnemonic symbolism — but "it encodes language" is a long way from "we can read it."