The Indus Valley Civilisation left behind thousands of inscribed objects, mostly small stone seals, carrying sequences of signs that nobody can read. Whether those sequences are writing at all is disputed, and the question has resisted more than a century of attempts.

The civilisation flourished from roughly 2600 to 1900 BC across what is now Pakistan and northwestern India, with cities including Mohenjo-daro and Harappa notable for grid planning, standardised brick sizes, and elaborate drainage. Its inscriptions survive on seals, sealings, pottery, copper tablets, and a small number of larger objects.

An Indus seal impression showing an animal above a short sequence of signs. Thousands of such objects survive and none has been read with any accepted decipherment.
An Indus seal impression showing an animal above a short sequence of signs. Thousands of such objects survive and none has been read with any accepted decipherment.Credit: PHGCOM IndusValleySeals.JPG (CC BY-SA 3.0).

Three obstacles compound one another.

The texts are extremely short. The average inscription runs to about five signs and the longest known is seventeen. Decipherments generally need extended text to work with; five characters offer almost nothing to constrain a hypothesis.

There is no bilingual. Egyptian hieroglyphs fell to the Rosetta Stone and Linear B to a combination of statistical work and a lucky correspondence with Greek. No Indus text has ever been found beside the same content in a known script.

The underlying language is unknown. Even a correct phonetic reading would produce sounds in a language nobody has identified. Candidates proposed include Dravidian, Indo-Aryan, and a language family with no surviving descendants, and the question is entangled with modern political disputes about the origins of Indian civilisation, which has not helped the scholarship.

Mohenjo-daro. The civilisation that built these cities practised grid planning and standardised construction, which is part of why the absence of readable administrative records is so striking.
Mohenjo-daro. The civilisation that built these cities practised grid planning and standardised construction, which is part of why the absence of readable administrative records is so striking.Credit: Saqib Qayyum (CC BY-SA 3.0).

In 2004 Steve Farmer, Richard Sproat, and Michael Witzel argued that the Indus signs are not writing. Their case rests on the extreme brevity, the very large number of signs occurring only once, the absence of long texts anywhere in a civilisation that built at this scale, and sign frequency patterns they argue differ from those of known scripts. On their account the symbols are non-linguistic emblems marking clan, deity, ownership, or commodity, more like heraldry than script.

The counter-argument, advanced by Asko Parpola and others and supported by statistical work led by Rajesh Rao, is that the conditional entropy of Indus sign sequences resembles that of known written languages rather than that of non-linguistic symbol systems. Critics of that work reply that the statistical test does not discriminate as cleanly as claimed and that structured non-linguistic systems can produce similar profiles.

Neither side has prevailed. The dispute matters beyond the script itself, because if the symbols are not writing then a large, sophisticated, planned urban civilisation managed without literacy, which would be unusual and would say something about what writing is actually for.

The extent of the Indus Valley Civilisation at its mature phase. Its cities were planned and standardised across a very large area, which makes the question of whether it had writing consequential.
The extent of the Indus Valley Civilisation at its mature phase. Its cities were planned and standardised across a very large area, which makes the question of whether it had writing consequential.Credit: Avantiputra7 (CC BY-SA 3.0).