Why Can’t We Read the Indus Script?

A small steatite seal is sitting in a museum drawer somewhere in Delhi or Karachi or London. It’s about the size of a postage stamp, perhaps a bit bigger. On it, a person carved a row of symbols more than 4,000 years in the past: a fish, a jar, some vertical strokes, and a figure that looks as if it’s sitting cross-legged. Next to those symbols is an animal — usually called a unicorn, though it’s really just a stylised bull with one horn pointing forward. The seal was probably pressed into clay to mark a shipment of grain or cloth or beads. It did its job. Then it was lost, buried, forgotten, and dug up again inside the Nineteen Twenties by way of archaeologists who realised, with a jolt, that they had been searching at a written language nobody on Earth may want to examine.

That was a hundred years ago. We still can’t read it.

This isn’t for lack of trying. Hundreds of human beings — students, linguists, retired engineers, enthusiastic amateurs, and at the very least one cryptographer — have taken a crack at the Indus script. They’ve attempted to hyperlink it to Dravidian, to Sanskrit, to Munda, to Sumerian, to Elamite, to Chinese, and even to Easter Island’s rongorongo. The quantity of published decipherment attempts is now properly over a hundred. Not one has survived scrutiny. As the archaeologist Mortimer Wheeler put it with characteristic bluntness back in the 1960s, the attempts to read the script are “invalid.” That verdict hasn’t really changed.

So what’s the holdup? I want to walk through this carefully, because the reasons are more interesting than the usual “we haven’t found the Rosetta Stone” line. That’s real; however, it’s most effective at the beginning. The actual trouble is a tangle of missing evidence, bizarre information, and a few assumptions that could have been incorrect from the start.

First, What Are We Even Dealing With?

The Indus Valley Civilisation — every now and then called the Harappan Civilisation — was one of the superb urban cultures of the historical world. At its peak, around 2600 to 1900 BCE, it covered a massive bite of what’s now Pakistan and northwest India. Its towns, Mohenjo-daro and Harappa chief amongst them, had grid-like streets, protected drains, standardised bricks, and public buildings that suggest a quite sophisticated society. They traded with Mesopotamia. They had uniform weights and measures. And they had a script.

We have somewhere between five and six thousand objects inscribed with this script. Almost all of them are small: seals, sealings (clay impressions), tiny tablets, copper tools, pottery sherds, a few ivory rods, and a couple of stoneware bangles. The overwhelming majority are seals — square or rectangular pieces of steatite, usually two to four centimetres on a side, carved with a few signs and an animal motif.

The inscriptions themselves are painfully short. The median length is about four or five signs. The longest known text — a single seal from Mohenjo-daro — has just twenty-six signs. That’s it. Twenty-six. Compare that to the tens of thousands of lines of cuneiform we have from Mesopotamia or the thousands of Egyptian inscriptions. We aren’t working with books or letters or royal proclamations. We’re working with ancient versions of shipping labels and ID tags.

And that, right there, is the first real problem. You can’t crack a writing system from a pile of grocery receipts. You need context. You need long texts where signs repeat in different combinations, where you can see grammar at work, and where you can guess at meaning from structure. Four signs on a seal don’t give you that. It gives you a puzzle piece with no picture on the box.

The Rosetta Stone Problem (Yes, It’s Real)

I know the Rosetta Stone comparison is a cliché, but it’s a cliché because it’s true. When Champollion figured out Egyptian hieroglyphs in the 1820s, he had a slab with the same decree written in three scripts — hieroglyphs, Demotic, and Greek. He could read Greek. That gave him an anchor, a way to test guesses, and a crib sheet.

The Indus script has nothing like this. No bilingual text. No parallel inscription in a known language. No handy translation key left by some ancient scribe. The Mesopotamians mention an area called Meluhha, which most students suppose changed into the Indus place, and that they communicate about trading with Meluhhan traders. But no Indus textual content has ever been located alongside a Mesopotamian one. No one has dug up a tablet that says “fish = min” or “jar = kumbha” in two languages.

Without that external anchor, every decipherment attempt is basically a guess. You look at a sign, you think, “That looks like a fish,” and you decide it means “fish” — or maybe it means “water” or “river” or a sound like “ma” or a clan name or a god. There’s no way to test any of those hypotheses against anything solid. You’re just building a castle on sand.

We Don’t Know What Language It Is

Here’s a thing people often miss: even if we found a bilingual tomorrow, we’d still need to know what language the Indus people spoke. And we don’t.

This is a genuinely hard problem. The Indus Valley Civilisation was huge and diverse. It traded with people who spoke Dravidian languages, Indo-Aryan languages, Munda languages, and maybe others. Any of those could have been the language of the script. Or it could have been a language that’s completely extinct, with no descendants at all. Or — and this is where it gets weird — the script might not have encoded a spoken language at all.

The Dravidian hypothesis is probably the most popular. It was championed by Asko Parpola in Finland and Iravatham Mahadevan in India, both of whom spent decades on it. They point to the fact that Dravidian languages are still spoken in South India and parts of Pakistan (Brahui), and they argue that many Indus signs look like Dravidian words for things — fish, for example, could be mīn, which also means “star” in some Dravidian languages, suggesting a rebus principle. It’s a clever argument. But it’s never been proven, and critics point out that you can find similar-looking wordplay in almost any language if you look hard enough.

The Indo-Aryan hypothesis has its own supporters, who argue that the script is related to early Sanskrit. The Munda hypothesis has a smaller but dedicated following. And then there’s the camp that says the whole thing might not be language at all — more on that in a minute.

The point is we don’t know. We have no Rosetta Stone, no known language, and no way to test any of these theories against hard evidence. It’s like trying to solve a crossword puzzle where you don’t know the language of the clues.

The Sign Count Is Weird

Here’s something that doesn’t get enough attention. The Indus script has somewhere between 400 and 600 distinct signs. That number is genuinely strange.

Think about it. Alphabets — like Latin or Greek or Brahmi — have twenty to forty signs. That’s because they represent individual sounds. Syllabaries — like Japanese kana or Linear B — have maybe fifty to a hundred signs, because they represent syllables. Logo-syllabic scripts — like Egyptian hieroglyphs, Mayan glyphs, or cuneiform — have hundreds or thousands of signs, because they mix word-signs with syllable-signs.

Four to six hundred signs sit right in the middle. It’s too many for an alphabet. It’s too few for a mature logo-syllabic system. It’s not quite a syllabary either. What is it?

Some scholars think it might be a logo-syllabic script that never fully developed — a system in transition, maybe used mainly for administrative purposes rather than literature. Others think the sign count is inflated because we’re counting different versions of the same sign as separate signs (the technical term is “allographs”). If you collapse the allographs, the number might drop to around 300 or so. But even that is awkward.

The sign count alone doesn’t prove anything, but it’s another piece of evidence that the Indus script doesn’t fit neatly into any category we know. It’s its own thing, and we don’t have a good model for what that thing might be.

What If It’s Not “Writing” at All?

This is the uncomfortable question that a few scholars have been asking for decades. What if the Indus script isn’t actually a writing system in the way we usually mean that term? What if it’s something else — a system of symbols for marking goods, identifying owners, recording quantities, or performing rituals — but not a way of encoding spoken language?

The most famous version of this argument came from Steve Farmer, Richard Sproat, and Michael Witzel in a 2004 paper. They argued that the Indus script lacks the kind of structural features you’d expect from a true writing system — things like long texts, repeated phrases, evidence of grammar, and a clear relationship between sign frequency and linguistic patterns. They suggested it might be a “non-linguistic sign system” — more like heraldry or religious iconography than actual writing.

This idea is controversial. A lot of scholars hate it. But it’s not crazy. Think about it: if the seals were used to mark shipments, you don’t necessarily need full sentences. You need a symbol for the commodity, a symbol for the quantity, and maybe a symbol for the owner or the destination. That’s not language; that’s notation. It’s like the symbols on a modern shipping box — arrows, fragile warnings, and recycling codes. They mean something, but they don’t encode a spoken language.

If the Farmer-Sproat-Witzel hypothesis is right, then the whole enterprise of trying to “read” the Indus script might be misguided. You can’t read a shipping label. You can only interpret it. And interpretation requires context that we mostly don’t have.

The Texts Might Be the Wrong Kind of Texts

Let’s say, for the sake of argument, that the Indus script is a real writing system. Even then, we might be stuck — because the surviving texts might be the wrong kind of texts.

Think about what we have. Seals. Tablets. Tiny inscriptions on pottery. These are administrative objects. They were probably used for trade, for accounting, and for marking ownership. They might contain names, titles, commodity labels, and numbers. They probably don’t contain poetry, mythology, history, or philosophy. Those things might have been written on perishable materials — palm leaves, cloth, wood — that haven’t survived in the archaeological record.

This is a real problem. If the entire surviving corpus of Indus writing consists of administrative labels, then even a perfect decipherment might not give us anything interesting. We might learn that a seal says “Ten jars of oil, property of the overseer of the granary.” Great. That’s useful for understanding the economy. It’s not going to tell us what the Indus people believed about the afterlife or how they told stories.

We’re not just missing the Rosetta Stone. We might be missing the entire library.

The Script Barely Changed for a Thousand Years

Here’s another oddity. The Indus script is remarkably stable. For more or less six hundred to one thousand years, across a big geographic area, the Middle Ages and their simple arrangement barely changed. In most writing systems, signs evolve. They simplify, they drift, they get adapted by different scribes in different regions. The Indus script didn’t do that.

Some scholars see this as evidence that the script was a fixed, standardised system — maybe something used mainly for ritual or administrative purposes, not for everyday communication. If it was a living language written by lots of people in lots of places, you’d expect more variation. The fact that it’s so consistent suggests it was used by a relatively small group of specialists, or that it was more like a code than a language.

Again, this doesn’t prove anything on its own. But it’s another clue that the Indus script might not be what we think it is.

A Century of Failed Decipherments

I want to spend a moment on the history of attempts, because it’s genuinely fascinating and also a bit depressing. The first serious attempt was by L.A. Wadsley in 1925, just a year after the seals were widely publicised. He tried to read them as Sumerian, convinced that the Indus people were racially related to the Aryans. (This was the 1920s. Race science was unfortunately very popular.) His work was, to put it kindly, not well received.

Since then, there have been dozens of attempts. Some have been by serious scholars with deep knowledge of ancient languages. Others have been by enthusiastic amateurs with a theory and a lot of free time. They’ve tried to link the script to Dravidian, to Sanskrit, to Munda, to Proto-Elamite, to Hittite, to Chinese, to Polynesian. One guy even tried to link it to the Easter Island script rongorongo. (It didn’t work.)

The fact that there are so many conflicting theories is itself a sign of how little solid ground there is. If any one of these decipherments were correct, you’d expect it to convince at least a majority of scholars. Instead, they all get published, get some press, get criticised, and then get quietly forgotten. The field is littered with abandoned theories.

What About Computers?

In the last couple of decades, there’s been a lot of excitement about using computational methods to crack the script. Researchers have built large digital databases of Indus signs. They’ve used statistical tools to analyse patterns — how often signs appear, in what order, next to what other signs. A landmark study in 2009 by Rajesh Rao and colleagues found that the Indus script’s patterns were closer to those of spoken language than to non-linguistic symbol systems. That was a big deal. It suggested the script does encode language, which contradicted the Farmer-Sproat-Witzel hypothesis.

Since then, there’s been more work — identifying allographs, modelling positional rules, building machine-learning models to predict the next sign in a sequence. The results have been interesting, but they haven’t produced a decipherment. As one recent review put it, there have been “significant advancements in grapheme and motif detection” but “ongoing shortcomings in the domain of semantic decipherment.” In plain English: we’re getting better at seeing how the signs are arranged, but not at figuring out what they mean.

Maybe We’ve Been Asking the Wrong Question

Here’s where I land, after reading a lot of this literature. I think the most interesting shift in recent years is the growing willingness to question the assumption that the Indus script must be a phonetic writing system — that each sign stands for a sound, and that the goal is to read it like a sentence.

After a hundred years of phonetic failure, some scholars are suggesting that the Indus script might be primarily semantic or administrative — a system that encoded concepts, quantities, and identities rather than the sounds of a spoken language. If that’s true, then trying to “read” it like a sentence might be the wrong goal entirely. We might need to learn to interpret it, not read it.

That might sound like giving up. I don’t think it is. It’s a recognition that the past doesn’t always fit our categories. The Indus people might have used writing differently than the Egyptians or the Mesopotamians did. They might have used it for things we don’t have good models for. And if we want to understand them, we might need to stop assuming they did things the way we expect.

So, Why Can’t We Read It?

Because we don’t have a bilingual. Because we don’t know the language. Because the texts are too short. Because the sign count is weird. Because the surviving inscriptions might be the wrong kind of writing. Because the script barely changed for a thousand years. Because a hundred years of brilliant, stubborn people have tried and failed.

But here’s the thing. I don’t think the Indus script is a failure. I think it’s a reminder that the past is not a puzzle designed for us to solve. The people who made those seals weren’t leaving a message for future archaeologists. They were running a civilisation. They were marking shipments and recording transactions and doing the ordinary administrative work of a complex society. The fact that we can’t read their labels doesn’t mean they were mysterious. It means we’re looking at the mundane paperwork of a lost world, and we’ve forgotten how to read the paperwork.

The search will continue. New datasets are being built. New statistical models are being run. Someone, somewhere, is probably staring at a sign right now and thinking, “Wait. What if…” Maybe they’ll crack it. Maybe they won’t. But either way, the Indus script remains one of the great mysteries of the human story — a script we can hold in our hands, but can’t quite hear.

And honestly? That’s kind of beautiful. Not everything has to be solved. Some doors stay closed. Some seals stay unbroken. The past keeps its secrets, and we keep asking.