Five Thousand Indus Inscriptions and Nobody Can Read One

Around 2800 BCE, in cities like Harappa and Mohenjo-daro, scribes started carving small soapstone seals — most no bigger than a postage stamp — with a row of symbols above the carved figure of a bull or unicorn. By 1900 BCE the cities were collapsing and the writing stopped. About 5,000 inscriptions have surfaced since the 1924 excavations.

That sounds like a lot until you count the words. The average Indus inscription is five symbols long. The full corpus is shorter than the back of a paperback. By comparison, the surviving Egyptian record from the same era runs into the millions of glyphs.

There is no bilingual Indus text. The Rosetta Stone gave Champollion two hundred Egyptian lines paired with a Greek translation he could already read; the Behistun inscription let Henry Rawlinson cross-check Old Persian against Elamite and Babylonian. With Indus, every theory has to bootstrap from the symbols alone — what Asko Parpola, who has worked on the script at Helsinki since the 1960s, calls trying to read a book whose only page is the front cover.

Parpola's team has argued for forty years that the underlying language is Dravidian, partly on the rebus pun that 'fish' and 'star' are homophones in many Dravidian tongues, which would explain the recurring fish glyph. Others read it as Indo-Aryan, or as not a language at all — Steve Farmer published the heretical version in 2004. In January 2025, Tamil Nadu's chief minister put up a $1 million prize for whoever cracks it. As of this writing the money is still on the table.