x Abu Dhabi, UAE Friday 21 July 2017

Code unknown: the fierce argument over ancient Indian symbols

Cover story In India - where 4,000 year-old stories still inspire death threats - historians, mathematicians and nationalists are going to battle over a vanished civilisation's script. S Subramanian reports.

The statistical physicist Ronojoy Adhikari has found himself an unlikely player in a fierce political-historical debate over India's origins.
The statistical physicist Ronojoy Adhikari has found himself an unlikely player in a fierce political-historical debate over India's origins.

In India - where 4,000 year-old stories still inspire death threats - historians, mathematicians and nationalists are going to battle over an ancient civilisation's script. S Subramanian reports. In 1856, searching for stone to anchor the railway tracks they were building between Karachi and Lahore, William and John Brunton, engineers working for the East India Railway Company, followed the directions of local residents to the site of an old, ruined town. There, they found 93 miles of perfect, kiln-fired bricks - and discovered the remains of Harappa, one of the two chief cities of the Bronze Age civilisation in the Indus valley. The Harappan ruins had been known previously, discovered by various explorers rambling around present-day Pakistan. But in the course of meticulously picking apart the bricks, the Bruntons unearthed enough artefacts to attract the attention of archeologists; their continued excavations revealed a record of an ancient civilisation whose urban ruins were scattered all across the vast Indus river basin. The discovery of Harappa revised, in one stroke, existing theories of ancient Indian history. Until then, the earliest known Indians were believed to be the literate Hindus who lived by the Rig Veda in the Second millennium BC. Modern Hindus trace their origins to this "Vedic civilisation", whose language and religion were considered wholly indigenous to the subcontinent. The existence of a separate pattern of settlement, an advanced civilisation predating the Vedic era by a few hundred years, raised confusing - and politically charged - questions. If the Indus Valley peoples were not Hindus, who were they? And where, then, did the Hindus come from? The study of this subject, peaceable enough for most of the 20th century, has in the last couple of decades been hauled down from its ivory tower. Now it messes about in the less exalted realms of politics and nationalism, internet vitriol and death threats - "weird territory," one participant says wryly, "that graduate school never prepared me for." The latest cycle of white-hot argument, for instance, began in the unlikely precincts of last April's issue of Science, when a group of researchers ran algorithms on the symbols scratched onto Indus Valley relics and suggested they bore the signatures of a written script - strengthening the possibility that this society was literate, and perhaps evolved into the Vedic civilisation rather than being displaced by it. The paper earned the prompt disfavour of what might be called the opposing team - scientists who don't believe the symbols make up a script at all.

Ronojoy Adhikari, one of the paper's authors, is an earnest young scientist at the Institute of Mathematical Sciences in Chennai who is new enough to Indus Valley studies that the name-calling still alarms him. He works on statistical physics, which mulls the properties of systems with thousands of units - atoms in a gas, for example. His interest in history, such as it was before this paper, had been purely that of the amateur enthusiast. Adhikari's study of history in school must have been similar to mine. The Indus Valley civilisation first appeared in my curriculum when I was in the ninth grade, in the mid-1990s, and from those lessons I seem to recall only strings of keywords, memorised hurriedly for some long-gone examination. The civilisation flourished between 2600 and 1900BC. There was an important port in a town named Lothal, and there was a Great Bath at Mohenjo-Daro. Both Harappa and Mohenjo-Daro had paved streets and marvellous drainage. Both had also yielded a treasury of seals and potsherds, into which had been scratched both recognisable motifs, such as the humpbacked bull, as well as symbols that welcomed all manner of interpretation. I still remember one that looked like a bald stick man clutching what seemed to be a comb, wedged into my memory by the nature of its irony. I also remember photographs of the ruins in my textbook, the clean, straight lines of their architectures still visible. Even in the speckled black-and-white reproductions, I could sense their beige hue, baked pale and hard by four thousand summers. What we were not taught was who exactly lived in Harappa and Mohenjo-Daro and the other Indus Valley towns, and this was for the very good reason that nobody knew - or knows - that with any certainty.

Discussions of ancient India usually begin with an invocation of Friedrich Max Müller, the 19th-century German philologist who wedded himself to Sanskrit and translated the Rig Veda. Müller was among the first scholars to propose an Aryan precursor to Indian and European civilisations - not the blue-eyed, apple-cheeked heroes of Nazi bedtime stories, but a common ancestor of high culture. Müller located them in central Asia and surmised that they spoke a proto-Indo-European tongue; when they separated, he believed, one branch headed towards India, developing Sanskrit and bearing the standard of Aryan culture. "What ? do we find in that ancient Sanskrit literature and cannot find anywhere else?" Müller asked of an audience of Indian Civil Service candidates at Cambridge in 1882. "My answer is: We find there the Aryan man, whom we know in his various characters, as Greek, Roman, German, Celt, and Slave [Slav], in an entirely new character." In British India, Müller's theory had two distinct effects. "In one way, it contributed to nationalist sentiment, and I can understand that," says Romila Thapar, one of India's most eminent historians, a silver-haired lady of great drive and sophistication. "Nationalism required an epic narrative of India being a great and ancient civilisation," Thapar tells me as we sit in the living room of her New Delhi house, a few meters away from her gorgeous study - a vast steppe of desk, bookshelves of burnished wood, a flood of light. "This theory provided that narrative." The Aryan hypothesis also boosted Hindutva, the political philosophy of Hindu nationalism. Like a gardener with a hothouse plant, Hindutva has carefully nurtured a specific version of Indian history, one filled with praise for the superiority of the early Hindus. At first, Müller's theory appeared to validate this perspective. Indeed, with a certain squint-eyed interpretation of the word "Aryan," it could even help explain the hierarchies of caste, suggesting that the upper-caste descendants of the Aryans were naturally positioned above the lower castes native to India. But this account also placed the possible birth of Hinduism, of Sanskrit, and of the Rig Veda outside India, which rubbed against the grain of the Hindutva credo that India had always been (and should always be) a Hindu nation. So, early in the 20th century, the Hindutva stance shifted, "from supporting the theory of an invasion," Thapar wrote in her book Early India, "to denying such an event, now arguing that the Aryans and their language, Sanskrit, were indigenous to India." Harappa and Mohenjo-Daro were built long before 1600BC, when the Aryans were thought to have arrived, so the Hindu nationalists concluded that there must have been no "invasion" - that these cities were already Aryan and Hindu. No act was too extreme in the defence of this official party line. It is commonly argued that, because the Rig Veda mentions horses and because no Indus Valley seal bears a horse motif, there is no continuity between the two societies - the horse, along with Sanskrit, must have come from outside India, after the Indus Valley civilisation. In response, a pair of scholars in the late 1990s were caught faking an image of a Harappan seal with a horse on it. All too inevitably, it came to be known as the Piltdown Horse. In India, even 4,000-year-old stories can be squeezed for some political capital. Over the years, various factions have gathered to skirmish over the identity of the "first Indians". All they needed was a ball to kick around, which is when the "script" rolled onto the field of play. If the Indus Valley symbols do not form a literate script, Sanskrit - or some proto-Sanskrit - almost certainly developed outside of India, and the Harappan civilisation was not a Hindu one. If they do, the Harappans were either the first Hindus - or they were literate Dravidians, the ancestors of today's south Indians, conquered and displaced by the Indo-Aryans. And so the script - or lack thereof - seems to hold the only hope of any sort of resolution to the politically loaded question: Who were the residents of these cities? And who, by extension, were the original Indians?

Adhikari's bare office, with two computers, a dry-erase board and little else, is situated on the fourth floor of a building that is either new or newly whitewashed. The computers are really all he needs for his work. On one of these two workstations, Adhikari ran his section of the Indus Valley algorithm; on their own computers, his Science collaborators - in Mumbai and at the University of Washington - ran theirs. The program worked for 45 minutes; interpreting the results took longer. "It's similar to speech recognition software," Adhikari says. "There are similar programs running in Google Translate, and in the predictive text feature on your cell phone." The central seam of the Science paper is a concept called conditional entropy, which can be explained like this: If you had reams of English text in front of you but didn't know the alphabet, you could still determine that the letter "Q" is nearly always followed by the letter "U," or that the letter "X" is rarely followed by the letter "S." According to this theory, if certain two-character sequences frequently appear - and other two-character sequences almost never do - then it is likely that the symbols encode language, and that you're looking at the spoor of a systematic script. "Our algorithm looked at a corpus of 417 Indus Valley symbols, from 2,906 individual texts," Adhikari says. "Then it ran every possible two-character combination to compute the probabilities." For comparison, the team also turned the program loose on Sumerian, English and Old Tamil scripts, and on two control batches of non-linguistic symbols. The results, plotted as a graph, look striking. The curve representing the Indus Valley symbols sits, along with the other known languages, in a dense bunch, which looks like a quartet of snakes trying to eat each other. The plots for the non-linguistic symbols, meanwhile, float serenely far above and below this intertwined bundle. On this metric, the Indus Valley curve coincides nearly exactly with that of Old Tamil, the most ancient of the known Dravidian languages. "When we published our results," Adhikari says, "we were immediately called Dravidian nationalists." Adhikari pulls up another graph, which plots a curve according to how often each symbol occurs, starts high and plunges downwards, corresponding to a pattern seen in other known languages. Finally, Adhikari pulls up a few Indus Valley "texts" - each one a discrete set of symbols - and asks me: "If you read from right to left, what do you see in common?" I peer at the screen. They all end in a symbol that looks like a U-shaped jar with elongated handles. "Is that a full stop of sorts?" "That's been suggested," Adhikari says. "Of the 417 symbols, the jar is by far the most common, and it is almost always at the end of a text." Some of these symbols, Adhikari and his team think, are definite text-ender signs and some are definite text-beginners - another signature, he says, of a script.

If it is a script at all, Indus Valley writing is neither syllabic, like English, where a letters stand in for specific sounds, nor logographic, like some Chinese characters, where there is an ideogram for every conceivable word. Instead it is thought to be "logosyllabic", a combination of the two - employing puns and rebuses that can only be deciphered within their context. To put English into a logosyllabic script, I'd draw something to signify the "sun" - a circle with rays streaming from it, say - but it could also mean "son," so you'd have to look at the next symbol to see if I'm talking about weather or family. Unless, of course, I choose to put an apostrophe modifier over the doodle, in which case, by a convention I am privy to but you are not, it could mean "rifle" or "volcano". To most of us, an ancient logosyllabic script can seem like a particular type of humorous anecdote, in that we had to have been there to really get it. His team's technique, Adhikari estimates, operates on an 80 per cent accuracy level; with some refinement, that could go up to 90 per cent. But the model comes with its flaws, and Adhikari is refreshingly willing to discuss them - again, I suspect, because he has entered the debate so recently. For one, the known texts are short to the point of laconic; the average length of a text is around five symbols, and the longest discovered text runs to 17 symbols. "So statistically, the margin of error is bound to go up," Adhikari admits. In contrast, some tablets of Linear B, the script of Mycenaean Greece that was deciphered in the 1950s, are jammed with row upon row of characters. The other weakness lies in the set of 417 symbols from which the algorithm plucked its combinations. The set, known as a concordance table, was assembled in 1977 by a Tamil epigraphist named Iravatham Mahadevan - India's most distinguished scholar of the Indus Valley inscriptions, and a staunch believer that the script is proto-Dravidian. But updates to the table have not kept pace with the symbols discovered since then. More troubling, according to Bryan Wells, a visiting scholar at the Institute of Mathematical Sciences who wrote his doctoral thesis on the Indus Valley symbols at Harvard, Mahadevan often condensed multiple symbols into one, casting out nuances that might be significant. Wells agrees, in theory, with Mahadevan and Adhikari - that the symbols form a script. "But I don't know why you need mathematical models to know it's a language," he says. "You just have to look at the damn thing."

In 1977, when Mahadevan built his concordance table - and indeed, for all of the 20th century - the Indus Valley inscriptions were widely regarded as a script. That was the case until 2004, when three scholars named Steve Farmer, Richard Sproat and Michael Witzel exploded the consensus with a strident paper called "The Collapse of the Indus-Script Thesis: The Myth of a Literate Harappan Civilization." Farmer, a comparative historian living in California, first entered the Indus Valley debate in 1999. "An outsider comes and looks at this, and it seems strange that everybody assumes it's a script," he says. Farmer and Witzel had wondered about the extreme brevity of the texts, and reject claims that longer manuscripts may have once existed on perishable materials that are now lost. Farmer likes to condense the main argument into a single sentence: "Not one ancient literate civilisation is known - including those that wrote routinely on perishable materials - that didn't also leave long texts behind on durable materials." Farmer, Sproat and Witzel offered a $10,000 prize to anybody who could find "just one Indus inscription that contains at least 50 symbols." The prize remains unclaimed. Farmer is unconvinced by the conditional entropy evidence of the Science paper; when I spoke to him, his team had just finished a draft of their response. "Their method shows that there's an order to the symbols, but that's true of every symbol system on the planet," he says. His favourite rebuttal cites the icons on motorway signboards. "You know that if you see a fork and a knife, you can get off the motorway and eat. Next to it, there's a bed, so you know there's a motel." These symbols appear in a fixed order on the signboards, Farmer points out, "but you can't call that a writing system. You can't sit down with these signs and have anybody dictate a speech to you." Perhaps unsurprisingly, "The Collapse of the Indus-Script Thesis" fired an instant uproar in India. "South Asians would come up to me after talks, some even crying, saying: 'You mean my ancestors were illiterate?'" he remembers. "I'd say: 'Listen, my name is Farmer, my ancestors were illiterate a hundred years ago!'" Farmer and Witzel were attacked as Westerners who couldn't accept the notion of an older, superior Asian culture. The wick of this discourse has burnt with a persistent heat. Farmer still receives indignant e-mails, one of which he shared with me in August. "Stevie how long even in the primitive and grotesque religion which spawned you, does it take for low life like you to evolve to a higher consciousness and morality?" the lady had written. "Isn't it a shame your religion is done with burning us at the stakes and genociding (sic) poor brown natives and enslaving the black?" Not all the abuse was so harmless: "We received a lot of death threats," Farmer says. "Even now, when I receive packages from India, I open them gingerly." He cites the rise of the Hindu right in India as the force behind the vitriol. During the Bharatiya Janata Party's term in power between 1999 and 2004, it seeded the history in school textbooks with its own preferred narrative. Suddenly there were great quantities of inflated detail about the rapacity of the Islamic invaders in India between the 11th and the 16th centuries, while the Indus Valley settlements were described as the original Vedic civilisation. When historians like Thapar, who had helped write the standard textbook in the mid-1960s, protested these amendments, they were promptly bullied. "I'd get calls at 2am saying: 'If you keep writing, we'll finish you off'," Thapar recalls. "My mother was alive at the time, and if she asked me who the caller was, I'd say it was a wrong number. She'd ask: 'Why do all these people dial the wrong number at 2am?'" Some scholars who find themselves on the same side of the debate as the Hindu right regard their incidental accomplices as something of a liability. "In every country, there is a lunatic fringe to the academic world, so what the Hindu right says is irrelevant," says Dilip Chakrabarti, an emeritus professor of south Asian archaeology at Cambridge University. Chakrabarti, a soft-voiced man with a self-effacing smile, converses almost entirely in questions; even an assertion will lilt upwards towards its end, adopting a Socratic, querying note. Sitting across a dining table in his New Delhi apartment, Chakrabarti dismantles theories for my benefit, like a mechanic who wants to show me exactly where the carburettor is clogged up. He dismisses the notion of the Aryan ancestor, calling it "a racist myth propagated as part of a colonial power game." He doesn't believe in the Indo-European language family, or the migration of a primitive Sanskrit south into India. "Aren't the very constructs of language families tenuous?" Chakrabarti asks. "And even if these families exist, how can you relate it to history or to a particular group of people called the Aryans?" In the Harappan culture, Chakrabarti spots many elements of what we recognise today as Hinduism. It could have produced the Rig Veda, Chakrabarti argues: there is no way to date the text, and it could easily have been written in 2200BC rather than 1200BC, the widely accepted date. Harappa may have possessed the horse, because some unearthed bones suggest as much, but Chakrabarti moans at the very mention of the horse conundrum. "The Rig Veda is a collection of hymns. Why should we expect in it an inventory of what is found in the culture that wrote it?" Later, he flips through a book he's written, India: An Archaeological History - published, ironically enough, by a firm called Aryan Books - and taps a particular section on the page. The passage describes a Harappan seal in which "a three-faced god with a horned headdress sits ? in yogic posture"; Chakrabarti unhesitatingly calls it a prototype of the Hindu deity Shiva. When I return home, I look up the seal. The central figure could be squatting in a posture that looks self-denying enough to be yogic, heels jammed into each other and thighs perfectly horizontal. He certainly wears an extravagantly curved headdress that resembles a pair of buffalo horns. Around him, a cordon of animals - an elephant, possibly a tiger and some cattle - is gathered. But I don't see the three faces, so I stare at the image, as if it's one of those pieces of magic art in which a spaceship suddenly emerges from a splash of pixels, and after a while, I've nearly convinced myself that this figure might be Shiva. Or it might not. Like the Indus Valley civilisation itself, the image on the seal seems ambiguous but accommodating, able to bear any meaning at all that is read into it.


S Subramanian, a regular contributor to The Review, is a journalist based in New Delhi.