The Indus Script Is Recorded in More Than 4,700 Inscriptions From About 60 Sites, and the Archaeological Survey of India Published Its Standard Catalogue of 417 Signs in 1977

More than 4,700 inscriptions in the Indus script have been dug up so far, according to a 2018 paper by Rajesh P. N. Rao. They sit on stamp seals, clay sealings, copper tablets, copper tools, ivory rods and pottery. The World History Encyclopedia says they come from some 60 excavation sites.

An Indian reference work organises them. In 1977 the Archaeological Survey of India (ASI) published I. Mahadevan’s The Indus Script: Texts, Concordance and Tables, a sign-by-sign index of the known texts that lists 417 signs. Farmer, Sproat and Witzel, who doubt that the signs record speech, call it the standard concordance.

About 60 sites, thousands of small objects

The first Harappan seal was found in 1872 to 1873, Farmer, Sproat and Witzel write. The Press Trust of India report on Tamil Nadu’s 2025 conference dates John Marshall’s announcement of the civilisation to 1924.

The objects are small. The seal impression that carries the longest text on a single surface measured 1 by 0.95 inches (2.54 by 2.41 cm), the 2004 paper says, as its excavators recorded it in the 1920s. The same authors count some 4,000 to 5,000 inscribed objects, on more than a dozen kinds of material that include shell, faience, bangles, tools and potsherds. Rao’s 2018 count of more than 4,700 inscriptions fits inside that range.

The script travelled. The Penn Museum’s 1944 bulletin, written by Leon Legrain, describes seals of Indian style from Kish, Ur, Umma and Lagash in Mesopotamia, with inscriptions in an unknown Indian script. It cites C. J. Gadd’s study of the Ur seals. The World History Encyclopedia adds that clay tags carrying the script have been found in Mesopotamia too.

The civilisation behind the script covered about a million square kilometres and peaked between 2600 and 1900 BCE, according to the 2010 paper by Yadav and colleagues. That is the civilisation of the Indus cities and their plumbing and of the dockyard town of Lothal.

Marks on pots before the cities

Dates for the first signs depend on what counts as writing. Jonathan Mark Kenoyer, who worked on the Harappa excavations, says the Ravi phase there (3800 to 2800 BCE) holds the earliest evidence of signs that were later used in the Indus script. He puts the start of the script proper in the Harappa phase, 2600 to 1900 BCE, with an “Early Indus Script” in wide use at Harappa and elsewhere during the Kot Diji phase before it.

Asko Parpola, whose proposed reading of the signs is Dravidian, draws the line later. He dates the earliest examples of Indus writing to the Kot Diji period, about 2800 to 2600 BCE at Harappa, and says most of the graffiti from that phase is not writing in the real sense. Nisha Yadav gives about 3300 BCE for pre-firing potters’ marks and post-firing graffiti. The World History Encyclopedia uses 3500 to 2700 BCE for single signs on Ravi and Kot Diji pottery.

Kenoyer’s summary article on Harappa.com traces the script to pre-firing and post-firing marks on pottery from sites that include Kalibangan and Balakot. He suggests the most common sign at the end of texts, Mahadevan’s sign 342, may appear in an earlier post-firing form on a Kot Diji fragment. Harappa.com’s introduction to the article says many steps in that development are still missing.

The World History Encyclopedia says the decline of the civilisation began by 1800 BCE and that writing disappeared. India then waited more than 1,000 years for writing to return, it adds.

What the 1977 concordance counts

The Roja Muthiah Research Library in Chennai catalogues Mahadevan’s book as published in 1977 by the Director General of the Archaeological Survey of India, with 829 pages. The record credits Mythili Rangarao for the computer programmes, N. Soundararajan for the composition software, R. Rajeswari as research assistant and Amarnath Sharma for the graphics.

Yadav and colleagues give its totals: 417 unique signs, 3,573 lines and 2,906 texts. Farmer, Sproat and Witzel count 13,372 individual sign occurrences across 2,905 objects, an average of 4.6 signs per object.

Other scholars count differently. Rao’s 2018 paper says Mahadevan reached the 417 by analysing the positions of signs in the corpus known around 1977, that Parpola used similar methods to estimate 386 signs in 1994, and that Wells has proposed 694. The 2004 authors say Wells’s earlier catalogue covers a different set of inscriptions and passes 600 signs because he classes far more variants as separate signs, a method they note has been widely criticised.

Five signs to a text

Texts are short. The World History Encyclopedia gives the average inscription as about five signs and the longest as 26. Farmer, Sproat and Witzel count 17 symbols as the longest on a single surface and say fewer than 1 in 100 inscriptions carry as many as 10. Yadav and colleagues report up to 14 signs on a line, with a mean text length of about five.

A few signs do most of the work. In Mahadevan’s concordance four of the 417 signs account for 21 per cent of the 13,372 occurrences, according to the 2004 paper. Yadav and colleagues find sign 342 is the most frequent in a filtered subset of 1,548 texts, at about 10 per cent of it, and that 69 signs cover roughly 80 per cent.

Rao says the writing runs mostly right to left, which means left to right on a seal and right to left on its impression. The World History Encyclopedia adds that some examples alternate direction from line to line.

Readings proposed, and no bilingual

No bilingual text, one that sets the script beside a known language, has been found. Rao lists “the lack of bilinguals, the brevity of the inscriptions” among the obstacles to decipherment, and writes of the many claimed decipherments that “none has been widely accepted by the community”.

The World History Encyclopedia says Indo-European and Dravidian are the language families most often proposed, with Austroasiatic, Sino-Tibetan and a lost family also suggested. Yadav and colleagues list an Indo-Aryan reading, a Dravidian reading by Parpola and a purely numerical reading by Subbarayappa, and say there is no consensus on any of them. Rao and co-authors wrote in 2009 that many prominent efforts had converged on a proto-Dravidian hypothesis.

Farmer, Sproat and Witzel write that more than a hundred claimed decipherments have been printed since 1877. They list the languages and scripts that have been tied to the signs, from Sumerian and Egyptian to early Indo-Aryan and proto-Dravidian. The Scroll report on the 2025 prize also puts attempts at more than 100.

Whether it records speech

The 2004 paper argues that the signs were a non-linguistic symbol system, like the emblems the authors cite from the ancient Near East. It points to the short texts, to high-frequency signs that rarely repeat inside one inscription, and to the lack of any long inscription.

A 2009 paper in Science by Rao and colleagues, including Mahadevan, answered with a statistical test. It found that the conditional entropy of Indus sign sequences, a measure of how freely one sign follows another, matches linguistic systems and sits far from the non-linguistic ones. The authors wrote that the results “increase the probability that the script represents language”.

Richard Sproat, one of the 2004 authors, replied in Language in 2014. The Linguistic Society of America’s release says his procedure classes the Indus symbols as non-linguistic, that he judges the earlier methods unfit for the purpose, and that corpus size, text length and text type all affect such statistics. He holds that showing a script is writing takes a credible decipherment or independent archaeological evidence of a literate culture.

Rao’s 2018 paper takes a third line. It studies miniature three-sided tablets that Richard Meadow and Kenoyer found at Harappa in 1997 and explores whether they were ration tokens, drawing parallels with proto-Elamite and proto-cuneiform records. Rao divides the corpus into short patterned texts and “complex” ones, such as the signboard at Dholavira, which our piece on Dholavira describes. Of his entropy results he writes that, while not proving a full-fledged writing system, they show script-like flexibility.

Kenoyer sets the Indus case beside the others. In his account early writing in Mesopotamia served accounting, in Egypt royal burial offerings, and in China contact with ancestors, while the Indus script stayed short and appears with trade, ideology, personal identification and common materials. He writes that it “clearly functioned very differently” from the writing of those societies.

A prize and a research chair in 2025

On 5 January 2025 Tamil Nadu’s Chief Minister announced a prize of US$1 million, about Rs 8.57 crore, for individuals or organisations that decipher the script, according to ThePrint’s Press Trust of India report. The answer has to be accepted by archaeologists. The announcement came at a three-day international conference in Chennai marking a hundred years since Marshall’s 1924 announcement, organised by the state Archaeology Department and the Indus Research Centre at the Roja Muthiah Research Library.

Scroll reports a Rs 2 crore grant for a research chair named after Iravatham Mahadevan, to be run by the department and the Indus Research Centre. ThePrint adds annual awards for two scholars and a foundation stone for a statue of John Marshall. As of 29 October 2026 the sources we read report no decipherment that archaeologists have accepted.