One Cliff Was Forgotten for a Thousand Years. On Another, They Carved a Temple Downward Out of a Single Rock

In April 1819, a British cavalry officer chasing a tiger through the hills of the western Deccan looked across a wooded ravine and saw an arch in the far cliff face where there should have been only rock.

His name was John Smith, of the 28th Madras Cavalry. A local boy led him down through the undergrowth and across the stream to the opening. Inside was a vaulted hall the length of a church, ribbed like the inside of a whale, with a stone stupa at the far end and the ghost of painted figures still clinging to the walls. Smith did what men of his century did at moments like this. He found a spot on the eleventh pillar and scratched his name and the date into a painting that was already more than a thousand years old.

That graffiti is still legible. It reads, more or less, “John Smith, 28th Cavalry, 28 April 1819.”

A forest had been keeping a secret

What Smith had walked into was Cave 10 at Ajanta, one of about thirty Buddhist halls cut into a horseshoe of cliff above the Waghora river in what is now Maharashtra. The cliff rises some 76 metres. The caves run along its inner curve like windows in a wall, and by the early nineteenth century the jungle had grown over the path so completely that from the valley floor you would never guess they were there.

It is tempting to call this a rediscovery, and the histories usually do. The honest version is narrower. The villages nearby had never lost the caves; herders sheltered in them, and the memory of them survived in the district. What happened in 1819 is that the caves re-entered the knowledge of the outside world, the world of colonial surveyors, museums, and copyists, which would spend the next century arguing over them and, in the process, wearing them down.

Because here is the uncomfortable part of the story. The paintings had survived a thousand years of abandonment in near darkness. They did not survive attention nearly so well. Early visitors lit fires inside, coated the murals with varnish and shellac to make them photograph better, and let copyists trace directly onto the surface. Two of the most ambitious sets of copies made in the nineteenth century were destroyed in fires before they could do much good. Ajanta’s real enemy was never the forest. It was the century that found it.

Two waves, six hundred years apart

Ajanta was not carved in one campaign. It was carved in two, separated by roughly six centuries of silence.

The first wave belongs to the 2nd to 1st century BCE, in the age of the Satavahana kings. The practice of cutting worship halls straight into living rock in India runs back to the masons of Ashoka in the century before, and Ajanta’s earliest caves follow that early Buddhist template. These are the austere ones: prayer halls, called chaityas, built around a stupa, and monastic dormitories, called viharas, with cells opening off a central square. In this early phase the Buddha is almost never shown as a human figure. He is present only as a symbol, a stupa, a wheel, an empty seat. Worship was aimed at what he represented, not at his face.

Then the site went quiet. For something like six hundred years, nobody added to it.

The second wave arrived in the late 5th century CE, under the Vakataka dynasty, and it was brief and extraordinary. The scholar Walter Spink argued for a chronology so tight that most of the great caves were begun, painted, and abandoned within roughly two decades, from about 460 to 480 CE, during and just after the reign of the Vakataka king Harishena. Spink’s compression is disputed, and plenty of specialists read the dating more loosely, but the broad picture holds: a sudden, concentrated burst of patronage produced the caves the world now travels to see. These were not funded by a single hand. Harishena’s court did some of the work, but ministers and subordinate lords paid for individual caves, their names recorded in dedicatory inscriptions on the walls they commissioned.

When that patronage collapsed after Harishena’s death, the work stopped, often mid-chisel. Some caves at Ajanta are frozen halfway through excavation, pillars roughed out but never finished, which is its own kind of gift: you can read the method in the unfinished stone.

Paint that outlived the empire that ordered it

The reason Ajanta matters out of all proportion to its size is the painting.

Very little wall painting survives anywhere from the ancient world. Ajanta is the great exception, the largest and finest body of early Buddhist painting left standing, and for centuries it was effectively the only window we had onto what classical Indian painting looked like at its height. The murals cover ceilings and walls in caves 1, 2, 16, and 17 above all, and they tell the Jataka tales, the stories of the Buddha’s previous lives, alongside scenes from his final life.

Strictly speaking they are not frescoes. A true fresco is painted into wet plaster. Ajanta’s artists worked on a dried surface, laying mineral pigments over a prepared ground of mud, dung, and lime, which is why the term you will sometimes see is tempera rather than fresco. The technique let them build up an astonishing density of figures: courts, processions, animals, lovers, ascetics, a whole crowded society rendered with a fluency of line that later Asian Buddhist art, from Sri Lanka to Central Asia to China, would inherit.

The single most reproduced image is the Bodhisattva Padmapani in Cave 1, a figure of compassion holding a blue lotus, eyes lowered, head tilted, wearing a jewelled crown that says more about the courtly taste of his patrons than about any renouncer’s life. Next to him, traditionally, stands Vajrapani. Between the two of them they have become the face of Ajanta, and of a whole lost tradition of Indian painting.

Sixty miles away, they went one better

Drive roughly a hundred kilometres southwest, toward the city of Aurangabad (now renamed Chhatrapati Sambhajinagar), and you reach the second cliff: Ellora.

Ellora is younger than Ajanta and, if anything, stranger. Its 34 major caves run for more than two kilometres along a basalt ridge, and they were cut across roughly four centuries, from about 600 to 1000 CE. What sets Ellora apart is not just the scale but the company the monuments keep. The caves are Buddhist, Hindu, and Jain, and they sit side by side.

The rough sequence runs: caves 1 to 12 are Buddhist, the earliest of the three groups; caves 13 to 29 are Hindu; and caves 30 to 34 are Jain, the latest. UNESCO reads the site as an emblem of the religious tolerance of ancient India, three faiths carving their sanctuaries into the same rock over the same centuries, sometimes within earshot of one another. That reading can be romanticised, and the patronage behind the caves was tangled up with the ambitions of rival dynasties, the Kalachuris, the Chalukyas, and above all the Rashtrakutas. But the physical fact remains remarkable. Nowhere else are the three traditions written into a single cliff at this scale.

They started at the top

And then there is Cave 16.

The Kailasa temple, named for the Himalayan mountain where the god Shiva is said to dwell, is not really a cave at all. It is a complete Hindu temple, courtyards, gateway, shrines, life-size stone elephants, a tower over the sanctum, every element a real temple would have, except that none of it was built. It was carved. Masons took a sloping basalt hillside and cut it away until a temple was left standing in the hole they had dug.

The method is what stops you. A normal building goes up from the ground. Kailasa went down from the sky. The carvers began at the top of the cliff and worked downward, freeing the tower first and the base last, releasing the whole structure out of the rock as a single connected mass. There was no scaffolding, because there was nothing to scaffold. There was also no room for error. You cannot add stone back to a monolith. Every cut was final.

The scale is hard to hold in the head. Estimates of the rock removed run from about 150,000 to 200,000 tonnes, with some popular accounts pushing higher; the figure is an estimate, not a measured quantity, so treat the neat numbers with caution. The temple rises more than 30 metres from the courtyard floor. It is often said to cover about twice the ground area of the Parthenon in Athens and to stand half again as tall, a comparison worth citing with the same care as the tonnage. What is documented, from two royal charters, is the patron: the Rashtrakuta king Krishna I, who reigned from about 756 to 773 CE. It is generally credited to his reign.

What the rock can and cannot tell you

Ask how long Kailasa took, or how many people it took, and you leave the ground of evidence. The often-repeated claims, eighteen years, or seven thousand labourers over a century and a half, are folklore and estimate, not record. No inscription gives a timetable. The temple’s own unfinished neighbours suggest the work was iterative and long, but the tidy figures that circulate online are guesses dressed as data.

This is the honest shape of both sites. What we know for certain is smaller and stranger than the legends. We know the caves were cut by hand out of solid basalt with iron tools. We know Ajanta was worked in two phases split by six centuries, and that its second phase was a short, brilliant flare of Vakataka patronage. We know Ellora braided three religions into one ridge over four hundred years, and that Kailasa was released from the rock top down under a king named Krishna. The rest, the exact crews, the precise years, the daily grind of chisel on stone, is mostly lost.

What is left is the work itself, which needs no exaggeration. On one cliff, painters left the ancient world’s finest surviving murals and a forest kept them for a thousand years. On another, masons stood on top of a mountain and carved a temple down out of it. Both were done by people whose names we mostly do not have, working stone that does not forgive, and both are still there, waiting on the far side of the same ravine John Smith looked across when he should have been watching for the tiger.